CN108399362A - A kind of rapid pedestrian detection method and device - Google Patents

A kind of rapid pedestrian detection method and device Download PDF

Info

Publication number
CN108399362A
CN108399362A CN201810069322.XA CN201810069322A CN108399362A CN 108399362 A CN108399362 A CN 108399362A CN 201810069322 A CN201810069322 A CN 201810069322A CN 108399362 A CN108399362 A CN 108399362A
Authority
CN
China
Prior art keywords
network
layer
target
layers
candidate
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201810069322.XA
Other languages
Chinese (zh)
Other versions
CN108399362B (en
Inventor
林倞
尹森堂
张冬雨
王青
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sun Yat Sen University
Original Assignee
Sun Yat Sen University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sun Yat Sen University filed Critical Sun Yat Sen University
Priority to CN201810069322.XA priority Critical patent/CN108399362B/en
Priority to PCT/CN2018/095058 priority patent/WO2019144575A1/en
Publication of CN108399362A publication Critical patent/CN108399362A/en
Application granted granted Critical
Publication of CN108399362B publication Critical patent/CN108399362B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24137—Distances to cluster centroïds
    • G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
    • G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
    • G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
    • G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Biophysics (AREA)
  • Human Computer Interaction (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a kind of rapid pedestrian detection method and devices, and described method includes following steps：Step S1 builds the configurable depth model based on convolutional neural networks, learns the network parameter for structure using training sample, obtains the model for test process；Step S2, input test sample, the changing rule for being perceived domain using neural network by trained model is detected the target object within the scope of different scale using different middle layers, predict the block diagram of target object in image, the present invention perceives the changing rule in domain by using neural network, the target object within the scope of particular dimensions is detected using different middle layers, the relationship in perception domain and article size has preferably been adapted to, has effectively increased testing result.

Description

A kind of rapid pedestrian detection method and device
Technical field
The present invention relates to pedestrian detection technology fields, more particularly to a kind of embedded system based on deep learning Rapid pedestrian detection method and device.
Background technology
As a part for target detection in computer vision, pedestrian detection has important meaning in the application of real world Justice, with the decline of the maturation and memory technology cost of image acquisition technology, more and more video cameras are deployed in public field Institute, on the other hand, with automatic Pilot, the implementation of intelligent transportation, vehicle-mounted camera also produces the video resource of magnanimity.Tradition Artificial screening and processing, not only inefficiency, expends a large amount of manpower and materials, and may introduce some human factors, causes Some deviations.In recent years, deep learning obtains unprecedented breakthrough in computer vision field, and not only efficiency far wins manpower, Accuracy is in many fields also above the mankind.Therefore, the project that the method for efficiently using deep learning carries out pedestrian detection receives Concern.
People is one of most important target in video monitoring or automatic Pilot, and the top priority of pedestrian detection is exactly to identify The presence of human body, and corresponding markup information is provided.Since the picture quality captured in real world is irregular, for The detection of wisp, the object blocked is always the difficult point of pedestrian detection, and on the other hand, vehicle-mounted camera also often captures Some fuzzy images, there is also the objects that a large amount of similar pedestrians are not but pedestrian in such image.And specific to embedded System, since the strong large-scale neural network model of recognition capability is generally difficult to efficient operate in the limited insertion of computing resource In formula equipment, and be real-time for the application demand of embedded device, thus take into account Detection accuracy and efficiency be towards The most important thing of the quick pedestrian detection of embedded system.
Invention content
In order to overcome the deficiencies of the above existing technologies, one of present invention is designed to provide a kind of quick pedestrian detection side Method and device perceive the changing rule in domain by using neural network, using different middle layers within the scope of particular dimensions Target object is detected, and has preferably been adapted to the relationship in perception domain and article size, has been effectively increased testing result.
Another object of the present invention is to provide a kind of rapid pedestrian detection method and device, by adjusting and training VGG- 16 network obtains adapting to the squeeze VGG-16 networks that embedded system requires, and effectively reduces the parameter amount of network model And accelerate computational efficiency.
A further object of the present invention is to provide a kind of rapid pedestrian detection method and device, passes through the method pair deconvoluted The characteristic pattern of particular network layer is amplified, and enhances the detection to wisp, compared to the method for conventional pictures amplification, almost Do not increase video memory and calculation amount.
The further object of the present invention is to provide a kind of rapid pedestrian detection method and device, by using target object The region of 1.5 times of sizes increases to as background semantic feature in network, the detection for fuzzy objective and remote wisp, There is splendid performance.
In view of the above and other objects, the present invention proposes a kind of rapid pedestrian detection method, include the following steps：
Step S1 builds the configurable depth model based on convolutional neural networks, learns structure using training sample Network parameter, obtain for test process model；
Step S2, input test sample, the changing rule for being perceived domain using neural network by trained model are used Different middle layers is detected the target object within the scope of different scale, predicts the block diagram of target object in image.
Preferably, step S1 further comprises：
The configurable depth model based on convolutional neural networks of structure；
Input training sample；
Initialize every layer of weight connected and biasing in convolutional neural networks and its parameter, including network layer；
Using propagated forward algorithm and Back Propagation Algorithm, learns the network parameter for structure using training sample, that is, use In the model of test process.
Preferably, the described depth model includes multiple dimensioned target candidate network and target detection network, the target Candidate network proposes the otherness of feature based on convolutional neural networks different layers, is generated respectively to different scale target in middle layer The candidate block diagram of object；The target detection network carries out essence on the basis of the candidate block diagram that the target candidate network exports The classification and detection of refinement.
Preferably, the convolutional neural networks are folded and are formed by convolutional layer, down-sampled layer, up-sampling layer heap.The convolutional layer Refer to that convolution algorithm is carried out on two-dimensional space to the image or characteristic pattern of input, extracts stratification feature；It is described down-sampled Layer is operated using the max-pooling not being overlapped, which is used to extract shape and deviates constant feature, while reducing spy Figure size is levied, computational efficiency is improved；The up-sampling layer refers to deconvoluting on two-dimensional space to the characteristic pattern of input Operation, to increase the pixel of characteristic pattern.
Preferably, the depth model uses Squeeze VGG-16 convolutional neural networks as backbone network, described 12 layer Fire module layer of the Squeeze VGG-16 convolutional neural networks using conv1-1 layers and followed by are characterized extraction Network structure.
Preferably, the target candidate network is on the basis of the Squeeze VGG-16 convolutional neural networks, according to volume Lamination feature, in Fire9, Fire12, conv6 and pooling layers increased, generation network branches, to carry out different scale Detect the recurrence of the candidate frame of object.
Preferably, the target detection network presets object candidate area on the basis of the object candidate area Background semantic information of the picture region of multiple size as target, Fire9 layers of characteristic pattern is once up-sampled, as Enhance the information perceived to wisp, and background semantic information is consolidated with up-sampling information by the pondization of area-of-interest Determine the feature of size, increase by one layer of full articulamentum later, carries out the recurrence of classification and final candidate frame.
Preferably, the training sample includes the markup information of pedestrian area in rgb image data and image, hands-on Image data is the small patch cut according to pedestrian region.
Preferably, the Back Propagation Algorithm need to first find out the target block diagram and image realistic objective of forward-propagating prediction The loss function of block diagramThen its gradient to parameter W is acquired, the algorithm that gradient declines is used to update W to minimize Loss functionIt is assumed that middle layer has M branch that can export object candidate area, lmIndicate the loss function of branch m, αmTable Show lmThe weight of function, S={ S1, S2..., SMRefer to the target object of corresponding scale, then loss functionIt may be defined as：
In order to achieve the above objectives, the present invention also provides a kind of quick pedestrian detecting systems, including：
Training unit is learnt for building the configurable depth model based on convolutional neural networks using training sample Go out the network parameter of structure, obtains the model for test process；
Detection unit is used for input test sample, and the variation for being perceived domain using neural network by trained model is advised Rule is detected the target object within the scope of different scale using different middle layers, predicts the frame of target object in image Figure.
Compared with prior art, the method that a kind of rapid pedestrian detection method of the present invention and device use for reference compression network, is adjusted Whole and training VGG-16 network obtains adapting to the squeeze VGG-16 networks that embedded system requires, and effectively reduces network The parameter amount of model simultaneously accelerates computational efficiency；On the other hand, differ with article size for perception domain in traditional detection method The problem of cause, (i.e. neural net layer is deeper, and perception domain is bigger, suitable using the changing rule in neural network perception domain by the present invention Detect larger target object), the target object within the scope of particular dimensions is detected using different middle layers, more preferably Adaptation perception domain and article size relationship, effectively increase testing result；In addition, in order to enhance the inspection to wisp It surveys, the present invention is amplified the characteristic pattern of particular network layer using the method deconvoluted, compared to the side of conventional pictures amplification Method hardly increases video memory and calculation amount；In order to enhance the detection for fuzzy objective, on the characteristic pattern of this layer, mesh is used The region of mark 1.5 times of sizes of object increases to as background semantic feature in network, for fuzzy objective and remote wisp Detection, have splendid performance.
Description of the drawings
Fig. 1 is a kind of step flow chart of rapid pedestrian detection method of the present invention；
Fig. 2 is Squeeze VGG-16 neural network structure schematic diagrames in the specific embodiment of the invention；
Fig. 3 is the schematic diagram of Fire modules in the specific embodiment of the invention；
Fig. 4 is the structural schematic diagram of target candidate network in the specific embodiment of the invention；
Fig. 5 is the structural schematic diagram of target detection network in the specific embodiment of the invention；
Fig. 6 is the process schematic of quick pedestrian detection in the specific embodiment of the invention；
Fig. 7 is a kind of system architecture diagram of quick pedestrian detection device of the present invention；
Fig. 8 is the detail structure chart of training unit in the specific embodiment of the invention；
Fig. 9 is the detail structure chart of detection unit in the specific embodiment of the invention.
Specific implementation mode
Below by way of specific specific example and embodiments of the present invention are described with reference to the drawings, those skilled in the art can Understand the further advantage and effect of the present invention easily by content disclosed in the present specification.The present invention can also pass through other differences Specific example implemented or applied, details in this specification can also be based on different perspectives and applications, without departing substantially from Various modifications and change are carried out under the spirit of the present invention.
Fig. 1 is a kind of step flow chart of rapid pedestrian detection method of the present invention.As shown in Figure 1, the present invention is a kind of quickly Pedestrian detection method includes the following steps：
Step S1 builds the configurable depth model based on convolutional neural networks, learns structure using training sample Network parameter, obtain for test process model.In the specific embodiment of the invention, the depth model is by two sub- networks Composition：First sub-network is multiple dimensioned target candidate network, for extracting character features and providing candidate region, specifically Ground, the target candidate network are proposed the otherness of feature based on convolutional neural networks different layers, are generated respectively to not in middle layer With the candidate block diagram of scale pedestrian；Second sub-network is target detection network, enhances the effect of detection, with target candidate Network share parameter, the classification and detection refined on the basis of candidate block diagram.Specifically, step S1 is further wrapped It includes：
Step S100 builds the configurable depth model based on convolutional neural networks.
The convolutional neural networks are folded and are formed by convolutional layer, down-sampled layer, up-sampling layer heap, and the convolutional layer refers to defeated The image or characteristic pattern entered carries out convolution algorithm on two-dimensional space, extracts stratification feature；The down-sampled layer uses The max-pooling operations not being overlapped, which is used to extract shape and deviates constant feature, while it is big to reduce characteristic pattern It is small, improve computational efficiency；The up-sampling layer refers to the behaviour deconvoluted on two-dimensional space to the characteristic pattern of input Make, to increase the pixel of characteristic pattern, is mainly used for target detection network, detection result is promoted, in the specific embodiment of the invention In, using Squeeze VGG-16 convolutional neural networks as backbone network, as shown in Fig. 2, the Squeeze VGG-16 convolution 12 layer Fire module of the neural network using conv1-1 layers and followed by are as convolutional layer, to extract feature；It is therein Pool1-pool5 is down-sampled layer；Using on ImageNet data sets advance trained model as initialization.That is this hair It is bright to train Squeeze VGG-16 as netinit in advance first with ImageNet data sets.
Fig. 3 is the structural schematic diagram of Fire modules in the specific embodiment of the invention.As shown in figure 3, Fire modules are by two The convolutional layer composition that the convolutional layer and a convolution kernel size that convolution kernel size is 1 × 1 are 3 × 3, it is therefore intended that with 1 × 1 volume Product core replaces 3 × 3 convolution kernel, to make parameter amount reduce 9 times, but in order to not influence the characterization ability of network, is not all of It substitutes, but a part is the convolution kernel with 1 × 1, a part uses 3 × 3 convolution kernel, another benefit done so is to subtract The input channel of few 3 × 3 convolution kernels, while the effect for reducing parameter amount is played, specifically, Fire modules use 1 × 1 before this Convolutional layer carries out dimensionality reduction operation to input layer, referring next to GoogLeNet structures, is extracted using 1 × 1 and 3 × 3 convolutional layer special Sign, finally connects two parts feature, such mode greatly reduces calculation amount and model parameter.
Fig. 4 is the configuration diagram of target candidate network in the specific embodiment of the invention.In the specific embodiment of the invention, The target candidate network is on the basis of Squeeze VGG-16 convolutional neural networks, according to convolutional layer feature, Fire9, Fire12, conv6 and increased pooling layers of 4 layers total, generation network branches, branch's progress different scale detect object The recurrence of the candidate frame of body.But for Fire-9 layers, the low layer of its relatively core network, compared to other layers to the shadow of gradient Ringing can be very big, and learning process is unstable, therefore more buffer (buffering) layer, as shown in det-conv layers in Fig. 4, Buffer layers avoid the gradient of detection branches from arriving trunk layer by direct back-propagated (backpropagation).
The present invention using neural network perception domain changing rule (i.e. neural net layer is deeper, perception domain it is bigger, be suitble to examine Survey larger target object), the target object within the scope of particular dimensions is detected using different middle layers, preferably The relationship for having adapted to perception domain and article size, effectively increases testing result.
Fig. 5 is the configuration diagram of target detection network in the specific embodiment of the invention.The target detection network and mesh Candidate network shared parameter is marked, the candidate frame of target candidate network is summarized, to enhance area of the monitoring network to object and background The ability of dividing.In the specific embodiment of the invention, the target detection network waits target on the basis of object candidate area Background semantic information of the picture region of 1.5 times of sizes of favored area as target；Fire9 layers of characteristic pattern adopted on primary Background semantic information is passed through the pond of area-of-interest by sample as the information that enhancing perceives wisp with up-sampling information (ROI pooling) obtains the feature of fixed size, increases by one layer of full articulamentum later, carries out time of classification and final candidate frame Return, specifically, the node of one proposals of cnn layers of connection of trunk, for summarizing the obtained candidate frame of target candidate network Information；On the other hand, the characteristic pattern for fire9 layers, W and H are the width and height for inputting picture, and cube 1 represents object The mapping in characteristic pattern in region, and cube 2 represents mapping of the regions context on characteristic pattern, the regions context are about 1.5 times of object area, while in order to reinforce the detection to wisp, then Fire9 layers are once up-sampled, Zhi Houyu Faster RCNN algorithms are similar, and the feature of fixed size is obtained using the pondization of area-of-interest；By Fire9 layers, treated Feature connect (concat) with the feature that proposals summarizes to together, increases by one layer of full articulamentum afterwards, carries out classification and final The recurrence of candidate frame, it will not be described here.
Step S101 inputs training sample.
Training process needs to provide the corresponding frame that personage is referred in image, while in order to accelerate to train, and training process will It cuts out to come from original image containing the image with reference to personage, forms patch (image block) one by one, patch is compared to original Beginning image smaller effectively accelerates training process to training.Specifically, in the present invention, the training sample of input includes The image data of the markup information of pedestrian area in rgb image data and image, hands-on is according to pedestrian region Cut obtained small patch (picture block).It is indicated with mathematical linguistics, training sampleWherein XiIt indicates One patch of training picture；In practical applications, in addition to this classification of pedestrian, also other classifications, for example, background, ride from The K classifications such as the people of driving vehicle, the people being seated, therefore labeled data Yi=(yi, bi) by class label yi∈ 0,1,2 ..., K } and block diagram coordinate pointsComposition, whereinFor the origin coordinates point in the block diagram upper left corner,For block diagram width and height.
Step S102 initializes every layer of weight connected and biasing in convolutional neural networks and its parameter, including network layer. Specifically, the present invention trains Squeeze VGG-16 convolutional neural networks initial as network in advance using ImageNet data sets Change.
Step S103 learns the network for structure using training sample using propagated forward algorithm and Back Propagation Algorithm Parameter is used for the model of test process.
In the present invention, the size normalization of input picture is first 3 × 480 × 640 by the propagated forward algorithm, The input as convolutional neural networks of patch and corresponding markup information for intercepting 3 × 448 × 448 sizes, by convolutional layer, Down-sampled layer and linear elementary layer (ReLU Nonlinearity Layer) is corrected, at Fire9 layers, characteristics of image figure size is 512×60×80；At Fire12 layers, characteristic pattern size is 512 × 30 × 40, and two branching characteristic figure sizes are successively later 512 × 15 × 20 and 512 × 8 × 10.On different characteristic figure, four coordinate points of target block diagram are obtained by the way of convolution And classification information, for Fire9 layers, it is assumed that only detection pedestrian and background, then it is 6 × 60 × 80 that output, which is characterized size, In 6 include background, two classifications of pedestrian and candidate four coordinate points of block diagram.In target detection network, each branch layer is obtained To candidate block diagram summarized in proposals nodes, while with Fire9 layers of background semantic information and up-sampling information warp The obtained feature of pondization operation for crossing area-of-interest is overlapped, and does that last block diagram returns and classification returns.
In the present invention, the Back Propagation Algorithm needs first to find out the target block diagram of positive (i.e. preceding to) propagation forecast With the loss function of image realistic objective block diagramThen its gradient to parameter W is acquired, the algorithm declined using gradient W is updated to minimize loss functionIt is assumed that middle layer has M branch that can export the object candidate area (perception of M scale Domain can approximately detect all target objects in image), lmIndicate the loss function of branch m, αmIndicate lmThe power of function Weight, S={ S1, S2..., SMRefer to the target object of corresponding scale, then loss functionIt may be defined as：
The loss function, for specific detection layers m, only target scale is in the range of m can be detected, just to damage It loses function to contribute, therefore loss function is defined as
Wherein, p (X)=(p0(X) ..., pK(X)) probability distribution of target category is indicated；λ is coefficient of balance；B is block diagram 4 coordinate points,Refer to the coordinate points that propagated forward obtains；In loss function, defines classification using cross entropy loss function and return Return, i.e.,
Lcls(p (X), y)=- logy(P(X)) (3)
The recurrence of target block diagram, definition are carried out using smooth manhatton distance standard (smooth L1 criterion) It is as follows
Step S2, the changing rule for perceiving domain using neural network by trained model use different middle layers pair Target object within the scope of different scale is detected, and predicts the block diagram of target object in image (such as pedestrian).
Specifically, step S2 further comprises：
Step S200 is loaded into trained model；
Step S201, input test sample；
Step S202, using trained model, the changing rule that domain is perceived by neural network uses different centres Layer is detected the pedestrian within the scope of different scale, the block diagram of pedestrian in prognostic chart picture.Fig. 6 is in the specific embodiment of the invention The process schematic of quick pedestrian detection utilizes the target candidate network in model in Squeeze VGG-16 convolutional Neural nets On the basis of network, according to convolutional layer feature, in fire9, fire12, conv6 and increased pooling layers total 4 layers generation net Network branch carries out object candidate area (middle layer a, middle layer b, middle layer c) that different scale detects object；Then it utilizes Target detection network, on the basis of object candidate area, using the picture region of 1.5 times of sizes of object candidate area as target Background semantic information, Fire9 layers of characteristic pattern is once up-sampled, as the information that perceive to wisp of enhancing, general Background semantic information obtains the feature of fixed size with up-sampling information by the pondization of area-of-interest, and one layer of increase later is complete Articulamentum carries out the recurrence of classification and final candidate frame.Preferably, in step S202, also using the method deconvoluted to spy The characteristic pattern for determining network layer is amplified.
Pedestrian detection method proposed by the present invention uses for reference both sides evaluation index respectively：Average precision mAP and per second Frame number FPS.MAP be used for evaluate last detection zone and real goal personage region friendship and than the case where, in different friendships and compare The average value of lower precision ratio；FPS, mainly efficiency index refer to manageable number of pictures per second.
Fig. 7 is a kind of system architecture diagram of quick pedestrian detection device of the present invention.As shown in fig. 7, the present invention is a kind of quickly Pedestrian detection device, including：
Training unit 70 utilizes training sample for building the configurable depth model based on convolutional neural networks The network parameter of structure is practised out, the model for test process is obtained.In the specific embodiment of the invention, 70 structures of training unit The depth model built is made of two sub- networks：First sub-network is multiple dimensioned target candidate network, for extracting personage Feature simultaneously provides candidate region, and specifically, which proposes the difference of feature based on convolutional neural networks different layers Property, the candidate block diagram to different scale pedestrian is generated respectively in middle layer；Second sub-network is target detection network, enhancing The effect of detection, with target candidate network shared parameter, the classification and detection refined on the basis of candidate block diagram. Specifically, as shown in figure 8, training unit 70 further comprises：
Model construction unit 701, for building the configurable depth model based on convolutional neural networks.
The convolutional neural networks are folded and are formed by convolutional layer, down-sampled layer, up-sampling layer heap, and the convolutional layer refers to defeated The image or characteristic pattern entered carries out convolution algorithm on two-dimensional space, extracts stratification feature；The down-sampled layer uses The max-pooling operations not being overlapped, which is used to extract shape and deviates constant feature, while it is big to reduce characteristic pattern It is small, computational efficiency is improved, the up-sampling layer refers to the behaviour deconvoluted on two-dimensional space to the characteristic pattern of input Make, to increase the pixel of characteristic pattern.In the specific embodiment of the invention, made using Squeeze VGG-16 convolutional neural networks For backbone network.
In the specific embodiment of the invention, the target candidate network is on Squeeze VGG-16 convolutional neural networks basis On, according to convolutional layer feature, in fire9, fire12, conv6 and increased pooling layers of 4 layers total, generation network point Branch, branch carry out the recurrence that different scale detects the candidate frame of object.But for fire-9 layers, its relatively core network Low layer, can be very big compared to influence of other layers to gradient, learning process is unstable, therefore a more buffer (buffering) Layer, buffer layers avoid the gradient of detection branches from arriving trunk layer by direct back-propagated (backpropagation).
The target detection network and target candidate network shared parameter, the candidate frame of target candidate network is summarized, with Enhance separating capacity of the monitoring network to object and background.In the specific embodiment of the invention, the target detection network, in mesh On the basis of marking candidate region, using the picture region of 1.5 times of sizes of object candidate area as the background semantic information of target；It will Fire9 layers of characteristic pattern is once up-sampled, as the information that is perceived to wisp of enhancing, by background semantic information with above adopt Sample information by area-of-interest pondization obtain fixed size feature, later increase by one layer of full articulamentum, carry out classification with The recurrence of final candidate frame, specifically, the subnet of one proposal of cnn layers of connection of trunk, W and H are the width for inputting picture And height, cube 1 represents the pooling of object area, and cube 2 represents the pooling in the regions context, context Region is about 1.5 times of object area, while in order to reinforce the detection to wisp, then is once up-sampled to Fire9 layers, It is similar with faster RCNN algorithms later, the feature of fixed size is obtained using the pondization of area-of-interest, increases by one layer later Full articulamentum carries out the recurrence of classification and final candidate frame.
Training sample input unit 702, for inputting training sample.
Specifically, training sampleWherein XiIndicate a patch of training picture, labeled data Yi=(yi, bi) by class label yiWith block diagram coordinate pointsComposition.
Initialization unit 703, the power for initializing every layer of connection in convolutional neural networks and its parameter, including network layer Weight and biasing.Specifically, the present invention trains Squeeze VGG-16 convolutional neural networks to make in advance using ImageNet data sets For netinit.
Sample training unit 704 learns for using propagated forward algorithm and Back Propagation Algorithm using training sample The network parameter of structure is used for the model of test process.
In the present invention, the size normalization of input picture is first 3 × 480 × 640 by the propagated forward algorithm, The input as convolutional neural networks of patch and corresponding markup information for intercepting 3 × 448 × 448 sizes, by convolutional layer, Down-sampled layer and linear elementary layer (ReLU Nonlinearity Layer) is corrected, at Fire9 layers, characteristics of image figure size is 512×60×80；At Fire12 layers, characteristic pattern size is 512 × 30 × 40, and two branching characteristic figure sizes are successively later 512 × 15 × 20 and 512 × 8 × 10.On different characteristic figure, four coordinate points of target block diagram are obtained by the way of convolution And classification information, for Fire9 layers, it is assumed that only detection pedestrian and background, then it is 6 × 60 × 80 that output, which is characterized size, In 6 include background, two classifications of pedestrian and candidate four coordinate points of block diagram.In target detection network, each branch layer is obtained To candidate block diagram summarized in proposals nodes, while with Fire9 layers of background semantic information and up-sampling information warp The obtained feature of pondization operation for crossing area-of-interest is overlapped, and does that last block diagram returns and classification returns.
The Back Propagation Algorithm needs the target block diagram for first finding out forward-propagating prediction and image realistic objective block diagram Loss functionThen its gradient to parameter W is acquired, the algorithm that gradient declines is used to update W to minimize loss letter NumberIt is assumed that middle layer has M branch that can export object candidate area, (the perception domain of M scale can approximately detect All target objects in image), lmIndicate the loss function of branch m, αmIndicate lmThe weight of function, S={ S1, S2..., SMRefer to The target object of corresponding scale, then loss functionIt may be defined as：
The loss function, for specific detection layers m, only target scale is in the range of m can be detected, just to damage It loses function to contribute, therefore loss function is defined as
Wherein, p (X)=(p0(X) ..., pK(X)) it is the probability distribution of target category.In loss function, cross entropy is used Loss function defines classification recurrence, i.e.,
Lcls(p (X), y)=- logy(P(X))
The recurrence that target block diagram is carried out using smooth L1 criterion, is defined as follows
Detection unit 71 is used for input test sample, perceives the variation in domain using neural network by trained model Rule is detected the target object (such as pedestrian) within the scope of different scale using different middle layers, predicts mesh in image Mark the block diagram of object (such as pedestrian).
Specifically, as shown in figure 9, detection unit 71 further comprises：
Model is loaded into unit 710, for being loaded into trained model；
Test sample input unit 711 is used for input test sample；
Image prediction unit 712 is perceived by trained model using neural network for utilizing trained model The changing rule in domain is detected the pedestrian within the scope of different scale using different middle layers, the frame of pedestrian in prognostic chart picture Figure.Specifically, image prediction unit 712 is using the target candidate network in model, in Squeeze VGG-16 convolutional Neural nets On the basis of network, according to convolutional layer feature, in Fire9, Fire12, conv6 and increased pooling layers total 4 layers generation net Network branch carries out the object candidate area that different scale detects object；Then target detection network is utilized, in target candidate area On the basis of domain, using the picture region of 1.5 times of sizes of object candidate area as the background semantic information of target, by Fire9 layers Characteristic pattern once up-sampled, as the information that perceive to wisp of enhancing, by background semantic information with up-sample information The feature of fixed size is obtained by the pondization of area-of-interest, increases by one layer of full articulamentum later, carries out classification and is finally waited Select the recurrence of frame.
In conclusion the method that a kind of rapid pedestrian detection method of the present invention and device use for reference compression network, adjusts and instructs The network for practicing VGG-16 obtains adapting to the squeeze VGG-16 networks that embedded system requires, and effectively reduces network model Parameter amount simultaneously accelerates computational efficiency；On the other hand, for perceiving domain in traditional detection method and article size is inconsistent asks Topic, the present invention using neural network perception domain changing rule (i.e. neural net layer is deeper, perception domain it is bigger, be suitble to detection greatly The target object of some), the target object within the scope of particular dimensions is detected using different middle layers, is preferably adapted to The relationship in perception domain and article size, effectively increases testing result；In addition, in order to enhance the detection to wisp, this hair It is bright that the characteristic pattern of particular network layer is amplified using the method deconvoluted, compared to the method for conventional pictures amplification, almost Do not increase video memory and calculation amount；In order to enhance the detection for fuzzy objective, on the characteristic pattern of this layer, target object is used The region of 1.5 times of sizes increases to as background semantic feature in network, the detection for fuzzy objective and remote wisp, There is splendid performance.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.Any Field technology personnel can without violating the spirit and scope of the present invention, and modifications and changes are made to the above embodiments.Therefore, The scope of the present invention, should be as listed in the claims.

Claims (10)

1. a kind of rapid pedestrian detection method, includes the following steps：
Step S1 builds the configurable depth model based on convolutional neural networks, learns the net for structure using training sample Network parameter obtains the model for test process；
Step S2, input test sample, the changing rule for perceiving domain using neural network by trained model use difference Middle layer the target object within the scope of different scale is detected, predict the block diagram of target object in image.
2. a kind of rapid pedestrian detection method as described in claim 1, which is characterized in that step S1 further comprises：
The configurable depth model based on convolutional neural networks of structure；
Input training sample；
Initialize every layer of weight connected and biasing in convolutional neural networks and its parameter, including network layer；
Using propagated forward algorithm and Back Propagation Algorithm, learn the network parameter for structure using training sample, that is, is used to survey The model of examination process.
3. a kind of rapid pedestrian detection method as claimed in claim 2, which is characterized in that the described depth model includes more rulers The target candidate network of degree and target detection network, the target candidate network are based on convolutional neural networks different layers and propose feature Otherness, generate the candidate block diagram to different scale target object respectively in middle layer；The target detection network is described The classification and detection refined on the basis of the candidate block diagram of target candidate network output.
4. a kind of rapid pedestrian detection method as claimed in claim 3, it is characterised in that：The convolutional neural networks are by convolution Layer, down-sampled layer, up-sampling layer heap is folded forms, the convolutional layer refer to the image or characteristic pattern of input on two-dimensional space Convolution algorithm is carried out, stratification feature is extracted；The down-sampled layer is operated using the max-pooling not being overlapped, the operation For extracting shape and deviating constant feature, while characteristic pattern size is reduced, improves computational efficiency；The up-sampling layer is Refer to the operation deconvoluted on two-dimensional space to the characteristic pattern of input, to increase the pixel of characteristic pattern.
5. a kind of rapid pedestrian detection method as claimed in claim 4, it is characterised in that：The depth model uses Squeeze VGG-16 convolutional neural networks are used as backbone network, the Squeeze VGG-16 convolutional neural networks The conv1-1 layers of network structure that extraction is characterized with 12 layers of Fire module layers followed by.
6. a kind of rapid pedestrian detection method as claimed in claim 5, it is characterised in that：The target candidate network is described On the basis of Squeeze VGG-16 convolutional neural networks, according to convolutional layer feature, in Fire9, Fire12, conv6 and increase Pooling layers, generate network branches, to carry out the recurrence that different scale detects the candidate frame of object.
7. a kind of rapid pedestrian detection method as claimed in claim 5, it is characterised in that：The target detection network is described On the basis of object candidate area, believe the picture region of object candidate area preset multiple size as the background semantic of target Breath, Fire9 layers of characteristic pattern is once up-sampled, and as the information that enhancing perceives wisp, and background semantic is believed Breath obtains the feature of fixed size with up-sampling information by the pondization of area-of-interest, increases by one layer of full articulamentum later, into The recurrence of row classification and final candidate frame.
8. a kind of rapid pedestrian detection method as described in claim 1, it is characterised in that：The training sample includes RGB figures As the markup information of pedestrian area in data and image, the image data of hands-on is cut according to pedestrian region The small patch arrived.
9. a kind of rapid pedestrian detection method as described in claim 1, it is characterised in that：The Back Propagation Algorithm, needs elder generation Find out the loss function of the target block diagram and image realistic objective block diagram of propagated forward predictionThen it is acquired to parameter The gradient of W uses the algorithm that gradient declines to update W to minimize loss functionIt is assumed that middle layer has M branch that can export Object candidate area, lmIndicate the loss function of branch m, αmIndicate lmThe weight of function, S={ S1, S2..., SMRefer to corresponding ruler The target object of degree, then loss functionIt may be defined as：
10. a kind of quick pedestrian detecting system, including：
Training unit learns structure using training sample for building the configurable depth model based on convolutional neural networks The network parameter built obtains the model for test process；
Detection unit is used for input test sample, perceives the changing rule in domain using neural network by trained model and makes The target object within the scope of different scale is detected with different middle layers, predicts the block diagram of target object in image.
CN201810069322.XA 2018-01-24 2018-01-24 Rapid pedestrian detection method and device Active CN108399362B (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201810069322.XA CN108399362B (en) 2018-01-24 2018-01-24 Rapid pedestrian detection method and device
PCT/CN2018/095058 WO2019144575A1 (en) 2018-01-24 2018-07-10 Fast pedestrian detection method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810069322.XA CN108399362B (en) 2018-01-24 2018-01-24 Rapid pedestrian detection method and device

Publications (2)

Publication Number Publication Date
CN108399362A true CN108399362A (en) 2018-08-14
CN108399362B CN108399362B (en) 2022-01-07

Family

ID=63094281

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810069322.XA Active CN108399362B (en) 2018-01-24 2018-01-24 Rapid pedestrian detection method and device

Country Status (2)

Country Link
CN (1) CN108399362B (en)
WO (1) WO2019144575A1 (en)

Cited By (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109089040A (en) * 2018-08-20 2018-12-25 Oppo广东移动通信有限公司 Image processing method, image processing apparatus and terminal device
CN109409364A (en) * 2018-10-16 2019-03-01 北京百度网讯科技有限公司 Image labeling method and device
CN109508675A (en) * 2018-11-14 2019-03-22 广州广电银通金融电子科技有限公司 A kind of pedestrian detection method for complex scene
CN109522855A (en) * 2018-11-23 2019-03-26 广州广电银通金融电子科技有限公司 In conjunction with low resolution pedestrian detection method, system and the storage medium of ResNet and SENet
CN109522966A (en) * 2018-11-28 2019-03-26 中山大学 A kind of object detection method based on intensive connection convolutional neural networks
CN109670439A (en) * 2018-12-14 2019-04-23 中国石油大学(华东) A kind of pedestrian and its location detection method end to end
CN109685718A (en) * 2018-12-17 2019-04-26 中国科学院自动化研究所 Picture quadrate Zoom method, system and device
CN109886066A (en) * 2018-12-17 2019-06-14 南京理工大学 Fast target detection method based on the fusion of multiple dimensioned and multilayer feature
CN109902800A (en) * 2019-01-22 2019-06-18 北京大学 The method of multistage backbone network detection generic object based on quasi- Feedback Neural Network
CN109993712A (en) * 2019-04-01 2019-07-09 腾讯科技（深圳）有限公司 Training method, image processing method and the relevant device of image processing model
CN110110793A (en) * 2019-05-10 2019-08-09 中山大学 Binocular image fast target detection method based on double-current convolutional neural networks
CN110110783A (en) * 2019-04-30 2019-08-09 天津大学 A kind of deep learning object detection method based on the connection of multilayer feature figure
CN110580726A (en) * 2019-08-21 2019-12-17 中山大学 Dynamic convolution network-based face sketch generation model and method in natural scene
CN110909615A (en) * 2019-10-28 2020-03-24 西安交通大学 Target detection method based on multi-scale input mixed perception neural network
CN111160527A (en) * 2019-12-27 2020-05-15 歌尔股份有限公司 Target identification method and device based on MASK RCNN network model
CN111176820A (en) * 2019-12-31 2020-05-19 中科院计算技术研究所大数据研究院 Deep neural network-based edge computing task allocation method and device
CN111247790A (en) * 2019-02-21 2020-06-05 深圳市大疆创新科技有限公司 Image processing method and device, image shooting and processing system and carrier
CN111242127A (en) * 2020-01-15 2020-06-05 上海应用技术大学 Vehicle detection method with granularity level multi-scale characteristics based on asymmetric convolution
CN111277751A (en) * 2020-01-22 2020-06-12 Oppo广东移动通信有限公司 Photographing method and device, storage medium and electronic equipment
CN111523351A (en) * 2019-02-02 2020-08-11 北京地平线机器人技术研发有限公司 Neural network training method and device and electronic equipment
CN111598951A (en) * 2020-05-18 2020-08-28 清华大学 Method, device and storage medium for identifying space target
CN111597945A (en) * 2020-05-11 2020-08-28 济南博观智能科技有限公司 Target detection method, device, equipment and medium
CN111709313A (en) * 2020-05-27 2020-09-25 杭州电子科技大学 Pedestrian re-identification method based on local and channel combination characteristics
CN111860508A (en) * 2020-07-28 2020-10-30 平安科技（深圳）有限公司 Image sample selection method and related equipment
CN112613472A (en) * 2020-12-31 2021-04-06 上海交通大学 Pedestrian detection method and system based on deep search matching
CN112613359A (en) * 2020-12-09 2021-04-06 苏州玖合智能科技有限公司 Method for constructing neural network for detecting abnormal behaviors of people
CN113379699A (en) * 2021-06-08 2021-09-10 上海电机学院 Transmission line insulator defect detection method based on deep learning
CN113486810A (en) * 2021-07-08 2021-10-08 国网江苏省电力有限公司徐州供电分公司 Intelligent identification method for birds stolen and hunted in park
CN111144203B (en) * 2019-11-19 2023-06-16 浙江工商大学 Pedestrian shielding detection method based on deep learning
CN111860508B (en) * 2020-07-28 2024-07-02 平安科技（深圳）有限公司 Image sample selection method and related equipment

Families Citing this family (274)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018176000A1 (en) 2017-03-23 2018-09-27 DeepScale, Inc. Data synthesis for autonomous control systems
US11157441B2 (en) 2017-07-24 2021-10-26 Tesla, Inc. Computational array microprocessor system using non-consecutive data formatting
US10671349B2 (en) 2017-07-24 2020-06-02 Tesla, Inc. Accelerated mathematical engine
US11409692B2 (en) 2017-07-24 2022-08-09 Tesla, Inc. Vector computational unit
US11893393B2 (en) 2017-07-24 2024-02-06 Tesla, Inc. Computational array microprocessor system with hardware arbiter managing memory requests
US11561791B2 (en) 2018-02-01 2023-01-24 Tesla, Inc. Vector computational unit receiving data elements in parallel from a last row of a computational array
US11215999B2 (en) 2018-06-20 2022-01-04 Tesla, Inc. Data pipeline and deep learning system for autonomous driving
US11361457B2 (en) 2018-07-20 2022-06-14 Tesla, Inc. Annotation cross-labeling for autonomous control systems
US11636333B2 (en) 2018-07-26 2023-04-25 Tesla, Inc. Optimizing neural network structures for embedded systems
US11562231B2 (en) 2018-09-03 2023-01-24 Tesla, Inc. Neural networks for embedded devices
JP2022504713A (en) 2018-10-11 2022-01-13 テスラ，インコーポレイテッド Systems and methods for training machine models with extended data
US11196678B2 (en) 2018-10-25 2021-12-07 Tesla, Inc. QOS manager for system on a chip communications
US11816585B2 (en) 2018-12-03 2023-11-14 Tesla, Inc. Machine learning models operating at different frequencies for autonomous vehicles
US11537811B2 (en) 2018-12-04 2022-12-27 Tesla, Inc. Enhanced object detection for autonomous vehicles based on field view
US11610117B2 (en) 2018-12-27 2023-03-21 Tesla, Inc. System and method for adapting a neural network model on a hardware platform
US11150664B2 (en) 2019-02-01 2021-10-19 Tesla, Inc. Predicting three-dimensional features for autonomous driving
US10997461B2 (en) 2019-02-01 2021-05-04 Tesla, Inc. Generating ground truth for machine learning from time series elements
US11567514B2 (en) 2019-02-11 2023-01-31 Tesla, Inc. Autonomous and user controlled vehicle summon to a target
US10956755B2 (en) 2019-02-19 2021-03-23 Tesla, Inc. Estimating object properties using visual image data
CN110659664B (en) * 2019-08-02 2022-12-13 杭州电子科技大学 SSD-based high-precision small object identification method
CN110633631B (en) * 2019-08-06 2022-02-18 厦门大学 Pedestrian re-identification method based on component power set and multi-scale features
CN110619268B (en) * 2019-08-07 2022-11-25 北京市新技术应用研究所 Pedestrian re-identification method and device based on space-time analysis and depth features
CN110533084B (en) * 2019-08-12 2022-09-30 长安大学 Multi-scale target detection method based on self-attention mechanism
CN110473195B (en) * 2019-08-13 2023-04-18 中山大学 Medical focus detection framework and method capable of being customized automatically
CN110427915B (en) * 2019-08-14 2022-09-27 北京百度网讯科技有限公司 Method and apparatus for outputting information
CN110705583B (en) * 2019-08-15 2024-03-15 平安科技（深圳）有限公司 Cell detection model training method, device, computer equipment and storage medium
CN110490252B (en) * 2019-08-19 2022-11-15 西安工业大学 Indoor people number detection method and system based on deep learning
CN110659576A (en) * 2019-08-23 2020-01-07 深圳久凌软件技术有限公司 Pedestrian searching method and device based on joint judgment and generation learning
CN110647816B (en) * 2019-08-26 2022-11-22 合肥工业大学 Target detection method for real-time monitoring of goods shelf medicines
CN110580727B (en) * 2019-08-27 2023-04-18 天津大学 Depth V-shaped dense network imaging method with increased information flow and gradient flow
CN110675309A (en) * 2019-08-28 2020-01-10 江苏大学 Image style conversion method based on convolutional neural network and VGGNet16 model
CN112446376B (en) * 2019-09-05 2023-08-01 中国科学院沈阳自动化研究所 Intelligent segmentation and compression method for industrial image
CN110728186B (en) * 2019-09-11 2023-04-07 中国科学院声学研究所南海研究站 Fire detection method based on multi-network fusion
CN110619365B (en) * 2019-09-18 2023-09-12 苏州经贸职业技术学院 Method for detecting falling water
CN110619676B (en) * 2019-09-18 2023-04-18 东北大学 End-to-end three-dimensional face reconstruction method based on neural network
CN110659601B (en) * 2019-09-19 2022-12-02 西安电子科技大学 Depth full convolution network remote sensing image dense vehicle detection method based on central point
CN110619309B (en) * 2019-09-19 2023-07-18 天地伟业技术有限公司 Embedded platform face detection method based on octave convolution and YOLOv3
CN110706239B (en) * 2019-09-26 2022-11-11 哈尔滨工程大学 Scene segmentation method fusing full convolution neural network and improved ASPP module
CN110674777A (en) * 2019-09-30 2020-01-10 电子科技大学 Optical character recognition method in patent text scene
CN110717903A (en) * 2019-09-30 2020-01-21 天津大学 Method for detecting crop diseases by using computer vision technology
CN110751076B (en) * 2019-10-09 2023-03-28 上海应用技术大学 Vehicle detection method
CN110781895B (en) * 2019-10-10 2023-06-20 湖北工业大学 Image semantic segmentation method based on convolutional neural network
CN110728238A (en) * 2019-10-12 2020-01-24 安徽工程大学 Personnel re-detection method of fusion type neural network
CN110728640B (en) * 2019-10-12 2023-07-18 合肥工业大学 Fine rain removing method for double-channel single image
CN111008554B (en) * 2019-10-16 2024-02-02 合肥湛达智能科技有限公司 Deep learning-based method for identifying pedestrians without giving away in dynamic traffic zebra stripes
CN111046723B (en) * 2019-10-17 2023-06-02 安徽清新互联信息科技有限公司 Lane line detection method based on deep learning
CN111008632B (en) * 2019-10-17 2023-06-09 安徽清新互联信息科技有限公司 License plate character segmentation method based on deep learning
CN110852179B (en) * 2019-10-17 2023-08-25 天津大学 Suspicious personnel invasion detection method based on video monitoring platform
CN110751644B (en) * 2019-10-23 2023-05-09 上海应用技术大学 Road surface crack detection method
CN111008562B (en) * 2019-10-31 2023-04-18 北京城建设计发展集团股份有限公司 Human-vehicle target detection method with feature map depth fusion
CN110826476A (en) * 2019-11-02 2020-02-21 国网浙江省电力有限公司杭州供电公司 Image detection method and device for identifying target object, electronic equipment and storage medium
CN110826552A (en) * 2019-11-05 2020-02-21 华中农业大学 Grape nondestructive automatic detection device and method based on deep learning
CN110826485B (en) * 2019-11-05 2023-04-18 中国人民解放军战略支援部队信息工程大学 Target detection method and system for remote sensing image
CN110837837B (en) * 2019-11-05 2023-10-17 安徽工业大学 Vehicle violation detection method based on convolutional neural network
CN111008567B (en) * 2019-11-07 2023-03-24 郑州大学 Driver behavior identification method
CN110852272B (en) * 2019-11-11 2023-03-28 上海应用技术大学 Pedestrian detection method
CN111461160B (en) * 2019-11-11 2023-07-14 天津津航技术物理研究所 Infrared imaging seeker target tracking method for preventing cloud and fog interference
CN111008994A (en) * 2019-11-14 2020-04-14 山东万腾电子科技有限公司 Moving target real-time detection and tracking system and method based on MPSoC
CN111222402A (en) * 2019-11-14 2020-06-02 北京理工大学 Crowd gathering density analysis method oriented to unmanned aerial vehicle image
CN111222534B (en) * 2019-11-15 2022-10-11 重庆邮电大学 Single-shot multi-frame detector optimization method based on bidirectional feature fusion and more balanced L1 loss
CN111126359B (en) * 2019-11-15 2023-03-28 西安电子科技大学 High-definition image small target detection method based on self-encoder and YOLO algorithm
CN110942008B (en) * 2019-11-21 2023-05-12 圆通速递有限公司 Deep learning-based face sheet information positioning method and system
CN110909797B (en) * 2019-11-22 2023-05-05 北京深睿博联科技有限责任公司 Image detection method and device, equipment and storage medium
CN110705540B (en) * 2019-11-25 2024-05-31 中国农业科学院农业信息研究所 Animal remedy production pointer instrument image recognition method and device based on RFID and deep learning
CN111105393B (en) * 2019-11-25 2023-04-18 长安大学 Grape disease and pest identification method and device based on deep learning
CN110956115B (en) * 2019-11-26 2023-09-29 证通股份有限公司 Scene recognition method and device
CN112949814B (en) * 2019-11-26 2024-04-26 联合汽车电子有限公司 Compression and acceleration method and device of convolutional neural network and embedded device
CN111046928B (en) * 2019-11-27 2023-05-23 上海交通大学 Single-stage real-time universal target detector and method with accurate positioning
CN111062278B (en) * 2019-12-03 2023-04-07 西安工程大学 Abnormal behavior identification method based on improved residual error network
CN111145195B (en) * 2019-12-03 2023-02-24 上海海事大学 Method for detecting portrait contour in video based on lightweight deep neural network
CN110986949B (en) * 2019-12-04 2023-05-09 日照职业技术学院 Path identification method based on artificial intelligence platform
CN111027449B (en) * 2019-12-05 2023-05-30 光典信息发展有限公司 Positioning and identifying method for paper archive electronic image archive chapter
CN110942144B (en) * 2019-12-05 2023-05-02 深圳牛图科技有限公司 Neural network construction method integrating automatic training, checking and reconstruction
CN111178148B (en) * 2019-12-06 2023-06-02 天津大学 Ground target geographic coordinate positioning method based on unmanned aerial vehicle vision system
CN110992238B (en) * 2019-12-06 2023-10-17 上海电力大学 Digital image tampering blind detection method based on dual-channel network
CN111008603B (en) * 2019-12-08 2023-04-18 中南大学 Multi-class target rapid detection method for large-scale remote sensing image
CN111160115B (en) * 2019-12-10 2023-05-02 上海工程技术大学 Video pedestrian re-identification method based on twin double-flow 3D convolutional neural network
CN111179338B (en) * 2019-12-10 2023-08-04 同济大学 Lightweight target positioning method for mobile power supply receiving end
CN111161217B (en) * 2019-12-10 2023-04-18 中国民航大学 Conv-LSTM multi-scale feature fusion-based fuzzy detection method
CN111062297B (en) * 2019-12-11 2023-05-23 青岛科技大学 Violent abnormal behavior detection method based on EANN deep learning model
CN111079642B (en) * 2019-12-13 2023-11-14 国网浙江余姚市供电有限公司 Line movable monitoring method and device and computer readable medium
CN110956157A (en) * 2019-12-14 2020-04-03 深圳先进技术研究院 Deep learning remote sensing image target detection method and device based on candidate frame selection
CN111178178B (en) * 2019-12-16 2023-10-10 汇纳科技股份有限公司 Multi-scale pedestrian re-identification method, system, medium and terminal combined with region distribution
CN111091101B (en) * 2019-12-23 2023-06-02 中国科学院自动化研究所 High-precision pedestrian detection method, system and device based on one-step method
CN111126310B (en) * 2019-12-26 2023-03-24 华侨大学 Pedestrian gender identification method based on scene migration
CN111178251B (en) * 2019-12-27 2023-07-28 汇纳科技股份有限公司 Pedestrian attribute identification method and system, storage medium and terminal
CN111161295B (en) * 2019-12-30 2023-11-21 神思电子技术股份有限公司 Dish image background stripping method
CN111160274B (en) * 2019-12-31 2023-03-24 合肥湛达智能科技有限公司 Pedestrian detection method based on binaryzation fast RCNN (radar cross-correlation neural network)
CN111199212B (en) * 2020-01-02 2023-04-07 西安工程大学 Pedestrian attribute identification method based on attention model
CN111209952B (en) * 2020-01-03 2023-05-30 西安工业大学 Underwater target detection method based on improved SSD and migration learning
CN111209860B (en) * 2020-01-06 2023-04-07 上海海事大学 Video attendance system and method based on deep learning and reinforcement learning
CN111259898B (en) * 2020-01-08 2023-03-24 西安电子科技大学 Crop segmentation method based on unmanned aerial vehicle aerial image
CN111259736B (en) * 2020-01-08 2023-04-07 上海海事大学 Real-time pedestrian detection method based on deep learning in complex environment
CN111275711B (en) * 2020-01-08 2023-04-07 西安电子科技大学 Real-time image semantic segmentation method based on lightweight convolutional neural network model
CN111260658B (en) * 2020-01-10 2023-10-17 厦门大学 Deep reinforcement learning method for image segmentation
CN111242010A (en) * 2020-01-10 2020-06-05 厦门博海中天信息科技有限公司 Method for judging and identifying identity of litter worker based on edge AI
CN111242839B (en) * 2020-01-13 2023-04-21 华南理工大学 Image scaling and clipping method based on scale level
CN111209887B (en) * 2020-01-15 2023-04-07 西安电子科技大学 SSD model optimization method for small target detection
CN113128316A (en) * 2020-01-15 2021-07-16 北京四维图新科技股份有限公司 Target detection method and device
CN111222519B (en) * 2020-01-16 2023-03-24 西北大学 Construction method, method and device of hierarchical colored drawing manuscript line extraction model
CN111259800A (en) * 2020-01-16 2020-06-09 天津大学 Neural network-based unmanned special vehicle detection method
CN111275688B (en) * 2020-01-19 2023-12-12 合肥工业大学 Small target detection method based on context feature fusion screening of attention mechanism
CN111275171B (en) * 2020-01-19 2023-07-04 合肥工业大学 Small target detection method based on parameter sharing multi-scale super-division reconstruction
CN111199220B (en) * 2020-01-21 2023-04-28 北方民族大学 Light-weight deep neural network method for personnel detection and personnel counting in elevator
CN111292366B (en) * 2020-02-17 2023-03-10 华侨大学 Visual driving ranging algorithm based on deep learning and edge calculation
CN111339871B (en) * 2020-02-18 2022-09-16 中国电子科技集团公司第二十八研究所 Target group distribution pattern studying and judging method and device based on convolutional neural network
CN111291820B (en) * 2020-02-19 2023-05-30 中国电子科技集团公司第二十八研究所 Target detection method combining positioning information and classification information
CN111428751B (en) * 2020-02-24 2022-12-23 清华大学 Object detection method based on compressed sensing and convolutional network
CN111368673B (en) * 2020-02-26 2023-04-07 华南理工大学 Method for quickly extracting human body key points based on neural network
CN111428567B (en) * 2020-02-26 2024-02-02 沈阳大学 Pedestrian tracking system and method based on affine multitask regression
CN113324864B (en) * 2020-02-28 2022-09-20 南京理工大学 Pantograph carbon slide plate abrasion detection method based on deep learning target detection
CN111339967B (en) * 2020-02-28 2023-04-07 长安大学 Pedestrian detection method based on multi-view graph convolution network
CN111339975B (en) * 2020-03-03 2023-04-21 华东理工大学 Target detection, identification and tracking method based on central scale prediction and twin neural network
CN111368726B (en) * 2020-03-04 2023-11-10 西安咏圣达电子科技有限公司 Construction site operation face personnel number statistics method, system, storage medium and device
CN111428586B (en) * 2020-03-09 2023-05-16 同济大学 Three-dimensional human body posture estimation method based on feature fusion and sample enhancement
CN111429410B (en) * 2020-03-13 2023-09-01 杭州电子科技大学 Object X-ray image material discrimination system and method based on deep learning
CN111461291B (en) * 2020-03-13 2023-05-12 西安科技大学 Long-distance pipeline inspection method based on YOLOv3 pruning network and deep learning defogging model
CN111460924B (en) * 2020-03-16 2023-04-07 上海师范大学 Gate ticket-evading behavior detection method based on target detection
CN111414909B (en) * 2020-03-16 2023-05-12 上海富瀚微电子股份有限公司 Target detection method and device
CN111368453B (en) * 2020-03-17 2023-07-07 创新奇智(合肥)科技有限公司 Fabric cutting optimization method based on deep reinforcement learning
CN111753625B (en) * 2020-03-18 2024-04-09 北京沃东天骏信息技术有限公司 Pedestrian detection method, device, equipment and medium
CN111462132A (en) * 2020-03-20 2020-07-28 西北大学 Video object segmentation method and system based on deep learning
CN111488805B (en) * 2020-03-24 2023-04-25 广州大学 Video behavior recognition method based on salient feature extraction
CN111563525A (en) * 2020-03-25 2020-08-21 北京航空航天大学 Moving target detection method based on YOLOv3-Tiny
CN111310861B (en) * 2020-03-27 2023-05-23 西安电子科技大学 License plate recognition and positioning method based on deep neural network
CN111310773B (en) * 2020-03-27 2023-03-24 西安电子科技大学 Efficient license plate positioning method of convolutional neural network
CN111414997B (en) * 2020-03-27 2023-06-06 中国人民解放军空军工程大学 Artificial intelligence-based method for battlefield target recognition
CN111460980B (en) * 2020-03-30 2023-04-07 西安工程大学 Multi-scale detection method for small-target pedestrian based on multi-semantic feature fusion
CN111462085B (en) * 2020-03-31 2023-09-19 上海大学 Digital image local filtering evidence obtaining method based on convolutional neural network
CN111553199A (en) * 2020-04-07 2020-08-18 厦门大学 Motor vehicle traffic violation automatic detection technology based on computer vision
CN111462108B (en) * 2020-04-13 2023-05-02 山西新华防化装备研究院有限公司 Machine learning-based head-face product design ergonomics evaluation operation method
CN111523645B (en) * 2020-04-16 2023-04-18 北京航天自动控制研究所 Convolutional neural network design method for improving small-scale target detection and identification performance
CN111597900B (en) * 2020-04-16 2023-10-24 浙江工业大学 Illegal dog walking identification method
CN111597897B (en) * 2020-04-16 2023-10-24 浙江工业大学 High-speed service area parking space recognition method
CN111695403B (en) * 2020-04-19 2024-03-22 东风汽车股份有限公司 Depth perception convolutional neural network-based 2D and 3D image synchronous detection method
CN111476314B (en) * 2020-04-27 2023-03-07 中国科学院合肥物质科学研究院 Fuzzy video detection method integrating optical flow algorithm and deep learning
CN111563440A (en) * 2020-04-29 2020-08-21 上海海事大学 Target detection method of multi-core iteration RPN based on heterogeneous convolution
CN111652846B (en) * 2020-04-30 2022-08-16 成都数之联科技股份有限公司 Semiconductor defect identification method based on characteristic pyramid convolution neural network
CN111597939B (en) * 2020-05-07 2023-04-18 西安电子科技大学 High-speed rail line nest defect detection method based on deep learning
CN111783685A (en) * 2020-05-08 2020-10-16 西安建筑科技大学 Target detection improved algorithm based on single-stage network model
CN111582452B (en) * 2020-05-09 2023-10-27 北京百度网讯科技有限公司 Method and device for generating neural network model
CN111783523B (en) * 2020-05-19 2022-10-21 中国人民解放军93114部队 Remote sensing image rotating target detection method
CN111709449B (en) * 2020-05-20 2023-08-18 西安理工大学 Multi-layer feature fusion small-scale target detection method based on clustering algorithm
CN112001878A (en) * 2020-05-21 2020-11-27 合肥合工安驰智能科技有限公司 Deep learning ore scale measuring method based on binarization neural network and application system
CN111881714B (en) * 2020-05-22 2023-11-21 北京交通大学 Unsupervised cross-domain pedestrian re-identification method
CN111626196B (en) * 2020-05-27 2023-05-16 西南石油大学 Knowledge-graph-based intelligent analysis method for body structure of typical bovine animal
CN111709311B (en) * 2020-05-27 2023-11-28 西安理工大学 Pedestrian re-identification method based on multi-scale convolution feature fusion
CN111832608B (en) * 2020-05-29 2023-09-12 上海海事大学 Iron spectrum image multi-abrasive particle identification method based on single-stage detection model yolov3
CN111652216B (en) * 2020-06-03 2023-04-07 北京工商大学 Multi-scale target detection model method based on metric learning
CN111652930B (en) * 2020-06-04 2024-02-27 上海媒智科技有限公司 Image target detection method, system and equipment
CN111709336B (en) * 2020-06-08 2024-04-26 杭州像素元科技有限公司 Expressway pedestrian detection method, equipment and readable storage medium
CN111881932B (en) * 2020-06-11 2023-09-15 中国人民解放军战略支援部队信息工程大学 FasterRCNN target detection algorithm for military aircraft
CN111860587B (en) * 2020-06-12 2024-02-02 长安大学 Detection method for small targets of pictures
CN111738124B (en) * 2020-06-15 2023-08-22 西安电子科技大学 Remote sensing image cloud detection method based on Gabor transformation and attention
CN111709935B (en) * 2020-06-17 2023-04-07 西安科技大学 Real-time coal gangue positioning and identifying method for ground moving belt
CN111797836B (en) * 2020-06-18 2024-04-26 中国空间技术研究院 Depth learning-based obstacle segmentation method for extraterrestrial celestial body inspection device
CN111723743A (en) * 2020-06-19 2020-09-29 北京邮电大学 Small-scale pedestrian rapid detection method
CN111832630A (en) * 2020-06-23 2020-10-27 成都恒创新星科技有限公司 Target detection method based on first-order gradient neural network
CN111784652B (en) * 2020-06-24 2024-02-06 西安电子科技大学 MRI (magnetic resonance imaging) segmentation method based on reinforcement learning multi-scale neural network
CN111767847B (en) * 2020-06-29 2023-06-09 佛山市南海区广工大数控装备协同创新研究院 Pedestrian multi-target tracking method integrating target detection and association
CN111814621B (en) * 2020-06-29 2024-01-23 中国科学院合肥物质科学研究院 Attention mechanism-based multi-scale vehicle pedestrian detection method and device
CN111832450B (en) * 2020-06-30 2023-11-28 成都睿沿科技有限公司 Knife holding detection method based on image recognition
CN111767878B (en) * 2020-07-03 2022-11-08 中国科学院自动化研究所 Deep learning-based traffic sign detection method and system in embedded device
CN112199983B (en) * 2020-07-08 2024-06-18 北京航空航天大学 Long-time large-range pedestrian re-identification method based on multi-level screening
CN111860265B (en) * 2020-07-10 2024-01-05 武汉理工大学 Multi-detection-frame loss balanced road scene understanding algorithm based on sample loss
CN111667030B (en) * 2020-07-13 2023-04-07 华东理工大学 Method, system and storage medium for realizing remote sensing image target detection based on deep neural network
CN111832479B (en) * 2020-07-14 2023-08-01 西安电子科技大学 Video target detection method based on improved self-adaptive anchor point R-CNN
CN111986149A (en) * 2020-07-16 2020-11-24 江西斯源科技有限公司 Plant disease and insect pest detection method based on convolutional neural network
CN111860637B (en) * 2020-07-17 2023-11-21 河南科技大学 Single-shot multi-frame infrared target detection method
CN111986126B (en) * 2020-07-17 2022-05-24 浙江工业大学 Multi-target detection method based on improved VGG16 network
CN111832513B (en) * 2020-07-21 2024-02-09 西安电子科技大学 Real-time football target detection method based on neural network
CN111881803B (en) * 2020-07-22 2023-10-31 安徽农业大学 Face recognition method based on improved YOLOv3
CN112001259A (en) * 2020-07-28 2020-11-27 联芯智能(南京)科技有限公司 Aerial weak human body target intelligent detection method based on visible light image
CN112036437B (en) * 2020-07-28 2024-06-07 农业农村部南京农业机械化研究所 Rice seedling detection model based on improved YOLOV network and method thereof
CN111915583B (en) * 2020-07-29 2024-02-09 西安电子科技大学 Vehicle and pedestrian detection method based on vehicle-mounted thermal infrared imager in complex scene
CN111985365A (en) * 2020-08-06 2020-11-24 合肥学院 Straw burning monitoring method and system based on target detection technology
CN112115291B (en) * 2020-08-12 2024-02-27 南京止善智能科技研究院有限公司 Three-dimensional indoor model retrieval method based on deep learning
CN111985464B (en) * 2020-08-13 2023-08-22 山东大学 Court judgment document-oriented multi-scale learning text recognition method and system
CN111986172B (en) * 2020-08-18 2024-06-04 华北电力科学研究院有限责任公司 Infrared image fault detection method and device for power equipment
CN111984879A (en) * 2020-08-19 2020-11-24 交控科技股份有限公司 User guiding method, device, equipment and storage medium applied to train
CN112001385B (en) * 2020-08-20 2024-02-06 长安大学 Target cross-domain detection and understanding method, system, equipment and storage medium
CN111985473A (en) * 2020-08-20 2020-11-24 中再云图技术有限公司 Method for identifying private business of store
CN111986186B (en) * 2020-08-25 2024-03-22 华中科技大学 Quantitative in-furnace PCB patch defect online detection system and method
CN112001339B (en) * 2020-08-27 2024-02-23 杭州电子科技大学 Pedestrian social distance real-time monitoring method based on YOLO v4
CN112364974B (en) * 2020-08-28 2024-02-09 西安电子科技大学 YOLOv3 algorithm based on activation function improvement
CN112149664B (en) * 2020-09-04 2024-05-07 浙江工业大学 Target detection method for optimizing classification and positioning tasks
CN112101434B (en) * 2020-09-04 2022-09-09 河南大学 Infrared image weak and small target detection method based on improved YOLO v3
CN112464765B (en) * 2020-09-10 2022-09-23 天津师范大学 Safety helmet detection method based on single-pixel characteristic amplification and application thereof
CN115661491A (en) * 2020-09-15 2023-01-31 重庆市农业科学院 Monitoring method for pest control in tea tree planting
CN112347843A (en) * 2020-09-18 2021-02-09 深圳数联天下智能科技有限公司 Method and related device for training wrinkle detection model
CN112163492B (en) * 2020-09-21 2023-09-08 华南理工大学 Long-time cross-scene optimization traffic object detection method, system and medium
CN112115885B (en) * 2020-09-22 2023-08-11 中国农业科学院农业信息研究所 Fruit tree fruiting branch shearing point positioning method based on deep convolutional neural network
CN112215100B (en) * 2020-09-27 2024-02-09 浙江工业大学 Target detection method for degraded image under unbalanced training sample
CN112200045B (en) * 2020-09-30 2024-03-19 华中科技大学 Remote sensing image target detection model establishment method based on context enhancement and application
CN112347851B (en) * 2020-09-30 2023-02-21 山东理工大学 Multi-target detection network construction method, multi-target detection method and device
CN112085126B (en) * 2020-09-30 2023-12-12 浙江大学 Single sample target detection method focusing on classification task
CN112183430B (en) * 2020-10-12 2024-04-05 河北工业大学 Sign language recognition method and device based on dual neural network
CN112232411B (en) * 2020-10-15 2024-05-14 苏州凌图科技有限公司 HarDNet-Lite optimization method in embedded platform
CN112419237B (en) * 2020-11-03 2023-06-30 中国计量大学 Deep learning-based automobile clutch master cylinder groove surface defect detection method
CN112381792B (en) * 2020-11-13 2023-05-23 中国人民解放军空军工程大学 Intelligent imaging on-line detection method for radar wave-absorbing coating/electromagnetic shielding film damage based on deep learning
CN112446308A (en) * 2020-11-16 2021-03-05 北京科技大学 Semantic enhancement-based pedestrian detection method based on multi-scale feature pyramid fusion
CN112396000B (en) * 2020-11-19 2023-09-05 中山大学 Method for constructing multi-mode dense prediction depth information transmission model
CN112434828B (en) * 2020-11-23 2023-05-16 南京富岛软件有限公司 Intelligent safety protection identification method in 5T operation and maintenance
CN112308062B (en) * 2020-11-23 2022-08-23 浙江卡易智慧医疗科技有限公司 Medical image access number identification method in complex background image
CN112580778A (en) * 2020-11-25 2021-03-30 江苏集萃未来城市应用技术研究所有限公司 Job worker mobile phone use detection method based on YOLOv5 and Pose-animation
CN112348036A (en) * 2020-11-26 2021-02-09 北京工业大学 Self-adaptive target detection method based on lightweight residual learning and deconvolution cascade
CN112487979B (en) * 2020-11-30 2023-08-04 北京百度网讯科技有限公司 Target detection method, model training method, device, electronic equipment and medium
CN112528826B (en) * 2020-12-04 2024-02-02 江苏省农业科学院 Control method of picking device based on 3D visual perception
CN112770325B (en) * 2020-12-09 2022-12-16 华南理工大学 Cognitive internet of vehicles spectrum sensing method based on deep learning
CN112560627A (en) * 2020-12-09 2021-03-26 江苏集萃未来城市应用技术研究所有限公司 Real-time detection method for abnormal behaviors of construction site personnel based on neural network
CN112396036B (en) * 2020-12-09 2023-08-08 中山大学 Method for re-identifying blocked pedestrians by combining space transformation network and multi-scale feature extraction
CN112633086B (en) * 2020-12-09 2024-01-26 西安电子科技大学 Near-infrared pedestrian monitoring method, system, medium and equipment based on multitasking EfficientDet
CN112560682A (en) * 2020-12-16 2021-03-26 重庆守愚科技有限公司 Valve automatic detection method based on deep learning
CN112465815B (en) * 2020-12-17 2023-09-19 杭州电子科技大学 Remote sensing target significance detection method based on edge main body fusion information
CN112651441B (en) * 2020-12-25 2022-08-16 深圳市信义科技有限公司 Fine-grained non-motor vehicle feature detection method, storage medium and computer equipment
CN112634367A (en) * 2020-12-25 2021-04-09 天津大学 Anti-occlusion object pose estimation method based on deep neural network
CN112699808B (en) * 2020-12-31 2024-06-07 深圳市华尊科技股份有限公司 Dense target detection method, electronic equipment and related products
CN112733848B (en) * 2021-01-08 2022-11-04 中国电子科技集团公司第二十八研究所 Target detection method based on multi-scale features and expanded inverse residual full-connection
CN112733714B (en) * 2021-01-11 2024-03-01 北京大学 VGG network-based automatic crowd counting image recognition method
CN112784921A (en) * 2021-02-02 2021-05-11 西北工业大学 Task attention guided small sample image complementary learning classification algorithm
CN112556682B (en) * 2021-02-07 2023-06-23 天津蓝鳍海洋工程有限公司 Automatic detection algorithm for underwater composite sensor target
CN112700444B (en) * 2021-02-19 2023-06-23 中国铁道科学研究院集团有限公司铁道建筑研究所 Bridge bolt detection method based on self-attention and central point regression model
CN112949508A (en) * 2021-03-08 2021-06-11 咪咕文化科技有限公司 Model training method, pedestrian detection method, electronic device and readable storage medium
CN112906718B (en) * 2021-03-09 2023-08-22 西安电子科技大学 Multi-target detection method based on convolutional neural network
CN113012208B (en) * 2021-03-22 2024-05-17 上海应用技术大学 Multi-view remote sensing image registration method and system
CN112906658A (en) * 2021-03-30 2021-06-04 航天时代飞鸿技术有限公司 Lightweight automatic detection method for ground target investigation by unmanned aerial vehicle
CN113312961A (en) * 2021-04-03 2021-08-27 国家计算机网络与信息安全管理中心 Logo recognition acceleration method
CN113221957B (en) * 2021-04-17 2024-04-16 南京航空航天大学 Method for enhancing radar information fusion characteristics based on center
CN113112511B (en) * 2021-04-19 2024-01-05 新东方教育科技集团有限公司 Method and device for correcting test paper, storage medium and electronic equipment
CN113076957A (en) * 2021-04-21 2021-07-06 河南大学 RGB-D image saliency target detection method based on cross-modal feature fusion
CN113011398A (en) * 2021-04-28 2021-06-22 北京邮电大学 Target change detection method and device for multi-temporal remote sensing image
CN113177545B (en) * 2021-04-29 2023-08-04 北京百度网讯科技有限公司 Target object detection method, target object detection device, electronic equipment and storage medium
CN113158968A (en) * 2021-05-10 2021-07-23 苏州大学 Embedded object cognitive system based on image processing
CN113408340B (en) * 2021-05-12 2024-03-29 北京化工大学 Dual-polarization SAR small ship detection method based on enhanced feature pyramid
CN113312995B (en) * 2021-05-18 2023-02-14 华南理工大学 Anchor-free vehicle-mounted pedestrian detection method based on central axis
CN113221787B (en) * 2021-05-18 2023-09-29 西安电子科技大学 Pedestrian multi-target tracking method based on multi-element difference fusion
CN113297961B (en) * 2021-05-24 2023-11-17 南京邮电大学 Target tracking method based on boundary feature fusion twin-cycle neural network
CN113222064A (en) * 2021-05-31 2021-08-06 苏州晗林信息技术发展有限公司 Image target object real-time detection method, system, terminal and storage medium
CN113343853B (en) * 2021-06-08 2024-06-14 深圳格瑞健康科技有限公司 Intelligent screening method and device for dental caries of children
CN113379709B (en) * 2021-06-16 2024-03-08 浙江工业大学 Three-dimensional target detection method based on sparse multi-scale voxel feature fusion
CN113449634A (en) * 2021-06-28 2021-09-28 上海翰声信息技术有限公司 Video detection method and device for processing under strong light environment
CN113379718B (en) * 2021-06-28 2024-02-02 北京百度网讯科技有限公司 Target detection method, target detection device, electronic equipment and readable storage medium
CN113469254B (en) * 2021-07-02 2024-04-16 上海应用技术大学 Target detection method and system based on target detection model
CN113449743B (en) * 2021-07-12 2022-12-09 西安科技大学 Coal dust particle feature extraction method
CN113642410B (en) * 2021-07-15 2024-03-29 南京航空航天大学 Method for detecting ampullaria gigas eggs based on multi-scale feature fusion and dynamic convolution
CN113361491A (en) * 2021-07-19 2021-09-07 厦门大学 Method for predicting pedestrian crossing intention of unmanned automobile
CN113657174A (en) * 2021-07-21 2021-11-16 北京中科慧眼科技有限公司 Vehicle pseudo-3D information detection method and device and automatic driving system
CN113487600B (en) * 2021-07-27 2024-05-03 大连海事大学 Feature enhancement scale self-adaptive perception ship detection method
CN113592825A (en) * 2021-08-02 2021-11-02 安徽理工大学 YOLO algorithm-based real-time coal gangue detection method
CN113591735A (en) * 2021-08-04 2021-11-02 上海新纪元机器人有限公司 Pedestrian detection method and system based on deep learning
CN113591854B (en) * 2021-08-12 2023-09-26 中国海洋大学 Low-redundancy rapid reconstruction method of plankton hologram
CN113805151A (en) * 2021-08-17 2021-12-17 青岛本原微电子有限公司 Attention mechanism-based medium repetition frequency radar target detection method
CN113869361A (en) * 2021-08-20 2021-12-31 深延科技(北京)有限公司 Model training method, target detection method and related device
CN113706491B (en) * 2021-08-20 2024-02-13 西安电子科技大学 Meniscus injury grading method based on mixed attention weak supervision migration learning
CN113989630B (en) * 2021-08-31 2024-04-23 中通服公众信息产业股份有限公司 Lens shielding judging method based on semantic analysis
CN113887330A (en) * 2021-09-10 2022-01-04 国网吉林省电力有限公司 Target detection system based on remote sensing image
CN113780193A (en) * 2021-09-15 2021-12-10 易采天成(郑州)信息技术有限公司 RCNN-based cattle group target detection method and equipment
CN113807243B (en) * 2021-09-16 2023-12-05 上海交通大学 Water obstacle detection system and method based on attention to unknown target
CN114067186B (en) * 2021-09-26 2024-04-16 北京建筑大学 Pedestrian detection method and device, electronic equipment and storage medium
CN113902024B (en) * 2021-10-20 2024-06-04 浙江大立科技股份有限公司 Small-volume target detection and identification method based on deep learning and dual-band fusion
CN113901944B (en) * 2021-10-25 2024-04-09 大连理工大学 Marine organism target detection method based on improved YOLO algorithm
CN115082909B (en) * 2021-11-03 2024-04-12 中国人民解放军陆军军医大学第一附属医院 Method and system for identifying lung lesions
CN114359644B (en) * 2021-12-22 2024-04-16 华南农业大学 Crop pest identification method based on improved VGG-16 network
CN114283320B (en) * 2021-12-25 2024-06-14 福州大学 Branch-free structure target detection method based on full convolution
CN114495166A (en) * 2022-01-17 2022-05-13 北京小龙潜行科技有限公司 Pasture shoe changing action identification method applied to edge computing equipment
CN114612769B (en) * 2022-03-14 2023-05-26 电子科技大学 Integrated sensing infrared imaging ship detection method integrated with local structure information
CN114884775A (en) * 2022-03-31 2022-08-09 南京邮电大学 Deep learning-based large-scale MIMO system channel estimation method
CN114863097B (en) * 2022-04-06 2024-05-31 北京航空航天大学 Infrared dim target detection method based on attention mechanism convolutional neural network
CN115019036B (en) * 2022-05-10 2024-02-27 西北工业大学 Small sample semantic segmentation method for learning non-target knowledge
CN115082386B (en) * 2022-06-07 2024-04-26 华南理工大学 Injection molding flaw detection method, device and medium based on normal sample auxiliary feature extraction
CN115423810B (en) * 2022-11-04 2023-03-14 国网江西省电力有限公司电力科学研究院 Blade icing form analysis method for wind generating set
CN116468928B (en) * 2022-12-29 2023-12-19 长春理工大学 Thermal infrared small target detection method based on visual perception correlator
CN116524293B (en) * 2023-04-10 2024-01-30 哈尔滨市科佳通用机电股份有限公司 Brake adjuster pull rod head loss fault identification method and system based on deep learning
CN117237614B (en) * 2023-11-10 2024-02-06 江西啄木蜂科技有限公司 Deep learning-based lake surface floater small target detection method

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105787439A (en) * 2016-02-04 2016-07-20 广州新节奏智能科技有限公司 Depth image human body joint positioning method based on convolution nerve network
CN105956608A (en) * 2016-04-21 2016-09-21 恩泊泰（天津）科技有限公司 Objective positioning and classifying algorithm based on deep learning
CN106934346A (en) * 2017-01-24 2017-07-07 北京大学 A kind of method of target detection performance optimization

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107341517B (en) * 2017-07-07 2020-08-11 哈尔滨工业大学 Multi-scale small object detection method based on deep learning inter-level feature fusion
CN107563349A (en) * 2017-09-21 2018-01-09 电子科技大学 A kind of Population size estimation method based on VGGNet

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105787439A (en) * 2016-02-04 2016-07-20 广州新节奏智能科技有限公司 Depth image human body joint positioning method based on convolution nerve network
CN105956608A (en) * 2016-04-21 2016-09-21 恩泊泰（天津）科技有限公司 Objective positioning and classifying algorithm based on deep learning
CN106934346A (en) * 2017-01-24 2017-07-07 北京大学 A kind of method of target detection performance optimization

Cited By (47)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109089040B (en) * 2018-08-20 2021-05-14 Oppo广东移动通信有限公司 Image processing method, image processing device and terminal equipment
CN109089040A (en) * 2018-08-20 2018-12-25 Oppo广东移动通信有限公司 Image processing method, image processing apparatus and terminal device
CN109409364A (en) * 2018-10-16 2019-03-01 北京百度网讯科技有限公司 Image labeling method and device
CN109508675A (en) * 2018-11-14 2019-03-22 广州广电银通金融电子科技有限公司 A kind of pedestrian detection method for complex scene
CN109522855A (en) * 2018-11-23 2019-03-26 广州广电银通金融电子科技有限公司 In conjunction with low resolution pedestrian detection method, system and the storage medium of ResNet and SENet
CN109522966B (en) * 2018-11-28 2022-09-27 中山大学 Target detection method based on dense connection convolutional neural network
CN109522966A (en) * 2018-11-28 2019-03-26 中山大学 A kind of object detection method based on intensive connection convolutional neural networks
CN109670439A (en) * 2018-12-14 2019-04-23 中国石油大学(华东) A kind of pedestrian and its location detection method end to end
CN109685718A (en) * 2018-12-17 2019-04-26 中国科学院自动化研究所 Picture quadrate Zoom method, system and device
CN109886066A (en) * 2018-12-17 2019-06-14 南京理工大学 Fast target detection method based on the fusion of multiple dimensioned and multilayer feature
CN109886066B (en) * 2018-12-17 2023-05-09 南京理工大学 Rapid target detection method based on multi-scale and multi-layer feature fusion
CN109902800A (en) * 2019-01-22 2019-06-18 北京大学 The method of multistage backbone network detection generic object based on quasi- Feedback Neural Network
CN109902800B (en) * 2019-01-22 2020-11-27 北京大学 Method for detecting general object by using multi-stage backbone network based on quasi-feedback neural network
CN111523351A (en) * 2019-02-02 2020-08-11 北京地平线机器人技术研发有限公司 Neural network training method and device and electronic equipment
CN111247790A (en) * 2019-02-21 2020-06-05 深圳市大疆创新科技有限公司 Image processing method and device, image shooting and processing system and carrier
CN109993712A (en) * 2019-04-01 2019-07-09 腾讯科技（深圳）有限公司 Training method, image processing method and the relevant device of image processing model
US11741581B2 (en) 2019-04-01 2023-08-29 Tencent Technology (Shenzhen) Company Limited Training method for image processing model, image processing method, network device, and storage medium
CN110110783A (en) * 2019-04-30 2019-08-09 天津大学 A kind of deep learning object detection method based on the connection of multilayer feature figure
CN110110793A (en) * 2019-05-10 2019-08-09 中山大学 Binocular image fast target detection method based on double-current convolutional neural networks
CN110580726A (en) * 2019-08-21 2019-12-17 中山大学 Dynamic convolution network-based face sketch generation model and method in natural scene
CN110580726B (en) * 2019-08-21 2022-10-04 中山大学 Dynamic convolution network-based face sketch generation model and method in natural scene
CN110909615B (en) * 2019-10-28 2023-03-28 西安交通大学 Target detection method based on multi-scale input mixed perception neural network
CN110909615A (en) * 2019-10-28 2020-03-24 西安交通大学 Target detection method based on multi-scale input mixed perception neural network
CN111144203B (en) * 2019-11-19 2023-06-16 浙江工商大学 Pedestrian shielding detection method based on deep learning
US11688163B2 (en) 2019-12-27 2023-06-27 Goertek Inc. Target recognition method and device based on MASK RCNN network model
WO2021129105A1 (en) * 2019-12-27 2021-07-01 歌尔股份有限公司 Mask rcnn network model-based target identification method and apparatus
CN111160527A (en) * 2019-12-27 2020-05-15 歌尔股份有限公司 Target identification method and device based on MASK RCNN network model
CN111176820A (en) * 2019-12-31 2020-05-19 中科院计算技术研究所大数据研究院 Deep neural network-based edge computing task allocation method and device
CN111242127B (en) * 2020-01-15 2023-02-24 上海应用技术大学 Vehicle detection method with granularity level multi-scale characteristic based on asymmetric convolution
CN111242127A (en) * 2020-01-15 2020-06-05 上海应用技术大学 Vehicle detection method with granularity level multi-scale characteristics based on asymmetric convolution
CN111277751B (en) * 2020-01-22 2021-06-15 Oppo广东移动通信有限公司 Photographing method and device, storage medium and electronic equipment
CN111277751A (en) * 2020-01-22 2020-06-12 Oppo广东移动通信有限公司 Photographing method and device, storage medium and electronic equipment
CN111597945B (en) * 2020-05-11 2023-08-18 济南博观智能科技有限公司 Target detection method, device, equipment and medium
CN111597945A (en) * 2020-05-11 2020-08-28 济南博观智能科技有限公司 Target detection method, device, equipment and medium
CN111598951A (en) * 2020-05-18 2020-08-28 清华大学 Method, device and storage medium for identifying space target
CN111598951B (en) * 2020-05-18 2022-09-30 清华大学 Method, device and storage medium for identifying space target
CN111709313A (en) * 2020-05-27 2020-09-25 杭州电子科技大学 Pedestrian re-identification method based on local and channel combination characteristics
CN111709313B (en) * 2020-05-27 2022-07-29 杭州电子科技大学 Pedestrian re-identification method based on local and channel combination characteristics
CN111860508A (en) * 2020-07-28 2020-10-30 平安科技（深圳）有限公司 Image sample selection method and related equipment
CN111860508B (en) * 2020-07-28 2024-07-02 平安科技（深圳）有限公司 Image sample selection method and related equipment
CN112613359A (en) * 2020-12-09 2021-04-06 苏州玖合智能科技有限公司 Method for constructing neural network for detecting abnormal behaviors of people
CN112613359B (en) * 2020-12-09 2024-02-02 苏州玖合智能科技有限公司 Construction method of neural network for detecting abnormal behaviors of personnel
CN112613472B (en) * 2020-12-31 2022-04-26 上海交通大学 Pedestrian detection method and system based on deep search matching
CN112613472A (en) * 2020-12-31 2021-04-06 上海交通大学 Pedestrian detection method and system based on deep search matching
CN113379699A (en) * 2021-06-08 2021-09-10 上海电机学院 Transmission line insulator defect detection method based on deep learning
CN113486810A (en) * 2021-07-08 2021-10-08 国网江苏省电力有限公司徐州供电分公司 Intelligent identification method for birds stolen and hunted in park
CN113486810B (en) * 2021-07-08 2024-06-18 国网江苏省电力有限公司徐州供电分公司 Intelligent identification method for park stolen hunting birds

Also Published As

Publication number Publication date
WO2019144575A1 (en) 2019-08-01
CN108399362B (en) 2022-01-07

Similar Documents

Publication Publication Date Title
CN108399362A (en) A kind of rapid pedestrian detection method and device
CN109993220B (en) Multi-source remote sensing image classification method based on double-path attention fusion neural network
CN109584248B (en) Infrared target instance segmentation method based on feature fusion and dense connection network
CN109493346B (en) Stomach cancer pathological section image segmentation method and device based on multiple losses
CN107220657B (en) A kind of method of high-resolution remote sensing image scene classification towards small data set
CN104809443B (en) Detection method of license plate and system based on convolutional neural networks
CN106897673B (en) Retinex algorithm and convolutional neural network-based pedestrian re-identification method
CN104834933B (en) A kind of detection method and device in saliency region
CN113065558A (en) Lightweight small target detection method combined with attention mechanism
CN110956094A (en) RGB-D multi-mode fusion personnel detection method based on asymmetric double-current network
CN108197606A (en) The recognition methods of abnormal cell in a kind of pathological section based on multiple dimensioned expansion convolution
CN104462494B (en) A kind of remote sensing image retrieval method and system based on unsupervised feature learning
CN108229425A (en) A kind of identifying water boy method based on high-resolution remote sensing image
CN107392925A (en) Remote sensing image terrain classification method based on super-pixel coding and convolutional neural networks
CN109271990A (en) A kind of semantic segmentation method and device for RGB-D image
CN111523521A (en) Remote sensing image classification method for double-branch fusion multi-scale attention neural network
CN108960404B (en) Image-based crowd counting method and device
CN113160062B (en) Infrared image target detection method, device, equipment and storage medium
CN103020265B (en) The method and system of image retrieval
CN108197669B (en) Feature training method and device of convolutional neural network
CN111291826A (en) Multi-source remote sensing image pixel-by-pixel classification method based on correlation fusion network
CN115116054B (en) Multi-scale lightweight network-based pest and disease damage identification method
CN108596818A (en) A kind of image latent writing analysis method based on multi-task learning convolutional neural networks
CN110222718A (en) The method and device of image procossing
CN114092833A (en) Remote sensing image classification method and device, computer equipment and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant