CN113762349A - Lightweight aliasing dense network classification method and system for marine organisms - Google Patents
Lightweight aliasing dense network classification method and system for marine organisms Download PDFInfo
- Publication number
- CN113762349A CN113762349A CN202110916982.9A CN202110916982A CN113762349A CN 113762349 A CN113762349 A CN 113762349A CN 202110916982 A CN202110916982 A CN 202110916982A CN 113762349 A CN113762349 A CN 113762349A
- Authority
- CN
- China
- Prior art keywords
- aliasing
- layer
- dense
- network
- marine organism
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 18
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 13
- 238000011176 pooling Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 230000004927 fusion Effects 0.000 abstract description 3
- 238000012549 training Methods 0.000 abstract description 3
- 241000251468 Actinopterygii Species 0.000 description 6
- 235000019580 granularity Nutrition 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005265 energy consumption Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
Images
Classifications
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a marine organism-oriented lightweight aliasing dense network classification method and system, which comprises the steps of obtaining marine organism images, loading the marine organism images into an aliasing dense network model which is established in advance and trained, and obtaining classification results; the aliasing dense network model is a convolutional neural network, a dense block is arranged in the convolutional neural network, the dense block comprises a plurality of aliasing network units, the aliasing network units are connected in a dense connection mode, the aliasing network units comprise a first grouping convolution layer, a batch regularization layer, a channel mixing layer, a depth separable convolution layer, a second grouping convolution layer and a serial connection layer which are sequentially connected, the serial connection layer is respectively connected with the input of the aliasing network units and the output of the second grouping convolution layer, and the output of the serial connection layer is connected with a linear rectification function. Compared with the prior art, the method has the advantages of acquiring more useful information, reducing the parameters of the model, realizing the fusion of the information, improving the network classification precision, accelerating the network training speed and the like.
Description
Technical Field
The invention relates to the field of image classification, in particular to a lightweight aliasing dense network classification method and system for marine organisms.
Background
Identification of marine fish species, habitat distribution, spawning site distribution, and the like are important research contents of marine organism science and marine environment science. Information science and technology has become an important means for marine biologists and environmentalists to conduct the above-mentioned research. The core problem is how to accurately classify marine fish targets based on the acquired visual data.
There are many fish in the ocean that have high visual similarity and it is difficult to distinguish them accurately without professional knowledge of marine biology.
Most of the existing methods for marine organism classification through a neural network have the defects of large calculation amount, high energy consumption, poor classification accuracy and the like.
Disclosure of Invention
The invention aims to provide a marine organism-oriented lightweight aliasing dense network classification method and system for overcoming the defects of large calculation amount, high energy consumption, poor classification accuracy and the like in the prior art.
The purpose of the invention can be realized by the following technical scheme:
a lightweight aliasing dense network classification method facing marine organisms comprises the steps of obtaining marine organism images to be classified, loading the marine organism images into an aliasing dense network model which is established in advance and trained, and obtaining classification results;
the aliasing dense network model is a convolutional neural network, a dense block is arranged in the convolutional neural network and comprises a plurality of aliasing network units, the aliasing network units are connected in a dense connection mode, the aliasing network units comprise a first grouping convolution layer, a batch regularization layer, a channel mixing layer, a depth separable convolution layer, a second grouping convolution layer and a series connection layer which are sequentially connected, the series connection layer is respectively connected with the input of the aliasing network units and the output of the second grouping convolution layer, and the output of the series connection layer is connected with a linear rectification function.
Further, the sizes of the convolution kernels of the first and second packet convolutional layers are both 1 × 1.
Further, the convolution kernel of the depth separable convolution layer has a size of 3 × 3 and a step size of 2, an average pooling layer is further arranged in a connection line between the concatenation layer and the input of the aliasing network unit, and the average pooling layer has a size of 3 × 3 and a step size of 2.
Further, the size of the convolution kernel of the depth separable convolution layer is 3 × 3, the step size is 1, and the cascade layer connects the inputs of the aliasing network units.
Further, the aliasing dense network model is provided with at least two symmetrically distributed dense blocks.
The invention also provides a marine organism-oriented lightweight aliasing dense network classification system, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the following steps: obtaining marine organism images to be classified, loading the marine organism images into a pre-established and trained aliasing dense network model, and obtaining a classification result;
the aliasing dense network model is a convolutional neural network, a dense block is arranged in the convolutional neural network and comprises a plurality of aliasing network units, the aliasing network units are connected in a dense connection mode, the aliasing network units comprise a first grouping convolution layer, a batch regularization layer, a channel mixing layer, a depth separable convolution layer, a second grouping convolution layer and a series connection layer which are sequentially connected, the series connection layer is respectively connected with the input of the aliasing network units and the output of the second grouping convolution layer, and the output of the series connection layer is connected with a linear rectification function.
Further, the sizes of the convolution kernels of the first and second packet convolutional layers are both 1 × 1.
Further, the convolution kernel of the depth separable convolution layer has a size of 3 × 3 and a step size of 2, an average pooling layer is further arranged in a connection line between the concatenation layer and the input of the aliasing network unit, and the average pooling layer has a size of 3 × 3 and a step size of 2.
Further, the size of the convolution kernel of the depth separable convolution layer is 3 × 3, the step size is 1, and the cascade layer connects the inputs of the aliasing network units.
Further, the aliasing dense network model is provided with at least two symmetrically distributed dense blocks.
Compared with the prior art, the invention has the following advantages:
(1) in order to acquire more useful information and reduce the parameters of the model as much as possible, separable convolution is introduced into a neural network for marine organism classification to increase the characteristic diagram and reduce the parameters of the model; in order to overcome the defect that in deep convolution, the input channels of each group only come from a part of channels of input information, which seriously affects the performance of a network model, a channel aliasing structure is introduced. After the deep separation convolution is obtained, each group is divided into smaller subgroups, and then the subgroups of each group are respectively sent to each group of the next deep convolution; by the method, the parameter quantity of the network is reduced, and meanwhile, the effective fusion of multi-channel information is improved.
(2) The invention adds a 1 x 1 grouping convolution at the head end and the tail end of a separable convolution and channel aliasing structure respectively to perform dimensionality reduction and recovery operation, thereby forming an aliasing network unit, forming a dense block by a plurality of aliasing network units in a dense connection mode, and constructing an aliasing dense network model by the dense block.
Drawings
FIG. 1 is a schematic structural diagram of two aliasing network units provided in the embodiment of the present invention;
fig. 2 is a schematic structural diagram of an aliasing dense network model provided in the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
In the embodiment, a marine fish classification method is explored from two granularities, namely coarse granularity classification and fine granularity classification, the families of the marine fish are determined by using a multi-class classification technology, and the categories of the marine fish are determined by using a fine granularity classification technology. In coarse-grained branch classification, channel aliasing and dense connection ideas are introduced to research a novel convolutional neural network model suitable for coarse classification, so that the network classification precision is improved, and the network training speed is accelerated.
In order to accelerate the training of the network and enable the network model to obtain a better prediction result under the conditions of poor computing power and low energy consumption, the embodiment provides a marine organism-oriented lightweight aliasing dense network classification method, which comprises the steps of obtaining marine organism images to be classified, loading the marine organism images into a pre-established and trained aliasing dense network model, and obtaining a classification result;
the aliasing dense network model is a convolutional neural network, a dense block is arranged in the convolutional neural network, the dense block comprises a plurality of aliasing network units, the aliasing network units are connected in a dense connection mode, the aliasing network units comprise a first grouping convolution layer, a batch regularization layer, a channel mixing layer, a depth separable convolution layer, a second grouping convolution layer and a serial connection layer which are sequentially connected, the serial connection layer is respectively connected with the input of the aliasing network units and the output of the second grouping convolution layer, and the output of the serial connection layer is connected with a linear rectification function.
The sizes of the convolution kernels of the first and second packet convolutional layers are both 1 × 1, and are used for performing dimensionality reduction and recovery operations, respectively.
The size of the convolution kernel of the depth separable convolution layer is 3 × 3; when the step length of the depth separable convolutional layer is 1, the tandem layer is directly connected with the input of the aliasing network unit; when the step size of the depth separable convolutional layer is 2, an average pooling layer is also arranged in a connecting line of the input of the concatenation layer and the aliasing network unit, the average pooling layer is 3 x 3 in size, and the step size is 2.
The above dense connection mode is specifically that the outputs of the preceding aliasing network units are all used as the inputs of the following aliasing network units.
Detailed concept description:
in this embodiment, separable convolution is introduced in order to obtain more useful information and to minimize the parameters of the model. Separable convolution is the addition of a 1 x 1 point-by-point convolution after the above-mentioned deep convolution, the addition of the feature map is achieved by this layer, and the parameters of the model are reduced. However, there is a serious problem in deep convolution: the input channels of each group are only from a part of the input information, which will seriously affect the performance of the network model. To solve this problem, a structure of channel aliasing is introduced. The structure, after obtaining the depth-separated convolution, divides each group into smaller subgroups and then sends the subgroups of each group separately to each group for the next depth convolution. In this way, fusion of information of different groups can be achieved.
Finally, the present embodiment combines the ideas of two structures of depth separable convolution and channel aliasing, and proposes an aliasing network element, as shown in fig. 1. Wherein, the left and right units respectively represent that different step sizes are adopted in the convolution process. Each unit starts and ends to perform dimensionality reduction and dimensionality restoration operations through 1 × 1 packet Convolution (Gconv), passes through a BN (Batch Normalization) layer, Channel aliasing (Channel Shuffle), Depth Separable Convolution (DWConv), and the like, and is finally concatenated with the input of each unit through a concatenation operation (Concat). A BN (Batch Normalization) layer is arranged after the 1 × 1 packet convolution at the beginning of each unit, the depth separable convolution layer and the 1 × 1 packet convolution at the end of each unit; the outputs after 1 × 1 packet convolution and concatenation operations at the beginning of each cell are provided with Relu activation functions. A plurality of aliasing network units form a Dense Block (Dense Block) in a Dense connection mode. Finally, an aliased dense network model is constructed by dense blocks, as shown in FIG. 2.
The aliasing dense network model is provided with at least two symmetrically distributed dense blocks.
In this embodiment, the aliasing dense network model is sequentially provided with a convolutional layer, a dense block, a convolutional layer, a max pooling layer … (combination of convolutional layer and max pooling layer), a dense block, a max pooling layer, and a full connection layer, and finally outputs a final prediction result.
The embodiment also provides a marine organism-oriented lightweight aliasing dense network classification system, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the marine organism-oriented lightweight aliasing dense network classification method.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.
Claims (10)
1. A marine organism-oriented lightweight aliasing dense network classification method is characterized by comprising the steps of obtaining marine organism images to be classified, loading the marine organism images into an aliasing dense network model which is built in advance and trained, and obtaining classification results;
the aliasing dense network model is a convolutional neural network, a dense block is arranged in the convolutional neural network and comprises a plurality of aliasing network units, the aliasing network units are connected in a dense connection mode, the aliasing network units comprise a first grouping convolution layer, a batch regularization layer, a channel mixing layer, a depth separable convolution layer, a second grouping convolution layer and a series connection layer which are sequentially connected, the series connection layer is respectively connected with the input of the aliasing network units and the output of the second grouping convolution layer, and the output of the series connection layer is connected with a linear rectification function.
2. The marine organism-oriented lightweight aliasing dense network classification method according to claim 1, wherein the sizes of convolution kernels of the first and second packet convolution layers are both 1 x 1.
3. The marine organism-oriented lightweight aliasing dense network classification method according to claim 1, wherein the convolution kernel of the depth separable convolution layer has a size of 3 x 3 and a step size of 2, an average pooling layer is further arranged in a connection line between the concatenation layer and an input of the aliasing network unit, and the average pooling layer has a size of 3 x 3 and a step size of 2.
4. The marine organism-oriented lightweight aliasing dense network classification method according to claim 1, wherein the convolution kernel of the depth separable convolution layer has a size of 3 x 3 and a step size of 1, and the cascade layer connects the inputs of the aliasing network units.
5. The marine organism-oriented lightweight aliasing dense network classification method as claimed in claim 1, wherein the aliasing dense network model is provided with at least two symmetrically distributed dense blocks.
6. A marine organism-oriented lightweight aliased dense network classification system comprising a memory and a processor, the memory storing a computer program, the processor invoking the computer program to perform the steps of: obtaining marine organism images to be classified, loading the marine organism images into a pre-established and trained aliasing dense network model, and obtaining a classification result;
the aliasing dense network model is a convolutional neural network, a dense block is arranged in the convolutional neural network and comprises a plurality of aliasing network units, the aliasing network units are connected in a dense connection mode, the aliasing network units comprise a first grouping convolution layer, a batch regularization layer, a channel mixing layer, a depth separable convolution layer, a second grouping convolution layer and a series connection layer which are sequentially connected, the series connection layer is respectively connected with the input of the aliasing network units and the output of the second grouping convolution layer, and the output of the series connection layer is connected with a linear rectification function.
7. The marine organism-oriented lightweight aliasing dense network classification system as claimed in claim 6, wherein the convolution kernels of the first and second packet convolution layers are both 1 x 1 in size.
8. The marine organism-oriented lightweight aliasing dense network classification system as claimed in claim 6, wherein the convolution kernel of the depth separable convolution layer has a size of 3 x 3 and a step size of 2, and an average pooling layer having a size of 3 x 3 and a step size of 2 is further provided in a connection line between the concatenation layer and an input of the aliasing network unit.
9. The marine organism-oriented lightweight aliased dense network classification system of claim 6 wherein the convolution kernels of the depth-separable convolutional layers have a size of 3 x 3 with a step size of 1, and the concatenation layer connects the inputs of the aliased network elements.
10. The marine organism-oriented lightweight aliasing dense network classification system as claimed in claim 6, wherein the aliasing dense network model is provided with at least two symmetrically distributed dense blocks.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110916982.9A CN113762349B (en) | 2021-08-11 | 2021-08-11 | Marine organism-oriented lightweight aliasing dense network classification method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110916982.9A CN113762349B (en) | 2021-08-11 | 2021-08-11 | Marine organism-oriented lightweight aliasing dense network classification method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113762349A true CN113762349A (en) | 2021-12-07 |
CN113762349B CN113762349B (en) | 2024-03-29 |
Family
ID=78788963
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110916982.9A Active CN113762349B (en) | 2021-08-11 | 2021-08-11 | Marine organism-oriented lightweight aliasing dense network classification method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113762349B (en) |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
IL239191A0 (en) * | 2015-06-03 | 2015-11-30 | Amir B Geva | Image classification system |
WO2018170393A2 (en) * | 2017-03-17 | 2018-09-20 | Portland State University | Frame interpolation via adaptive convolution and adaptive separable convolution |
CN109509149A (en) * | 2018-10-15 | 2019-03-22 | 天津大学 | A kind of super resolution ratio reconstruction method based on binary channels convolutional network Fusion Features |
WO2019079895A1 (en) * | 2017-10-24 | 2019-05-02 | Modiface Inc. | System and method for image processing using deep neural networks |
US20190369191A1 (en) * | 2018-05-31 | 2019-12-05 | The Board Of Trustees Of The Leland Stanford Junior University | MRI reconstruction using deep learning, generative adversarial network and acquisition signal model |
CN110570353A (en) * | 2019-08-27 | 2019-12-13 | 天津大学 | Dense connection generation countermeasure network single image super-resolution reconstruction method |
CN110956197A (en) * | 2019-10-28 | 2020-04-03 | 新绎健康科技有限公司 | Method and system for establishing pulse wave noise signal identification model based on convolutional neural network |
WO2020093042A1 (en) * | 2018-11-02 | 2020-05-07 | Deep Lens, Inc. | Neural networks for biomedical image analysis |
CN111723685A (en) * | 2020-06-01 | 2020-09-29 | 齐齐哈尔大学 | Remote sensing scene classification method based on branch feature fusion convolutional network |
-
2021
- 2021-08-11 CN CN202110916982.9A patent/CN113762349B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
IL239191A0 (en) * | 2015-06-03 | 2015-11-30 | Amir B Geva | Image classification system |
WO2018170393A2 (en) * | 2017-03-17 | 2018-09-20 | Portland State University | Frame interpolation via adaptive convolution and adaptive separable convolution |
WO2019079895A1 (en) * | 2017-10-24 | 2019-05-02 | Modiface Inc. | System and method for image processing using deep neural networks |
US20190369191A1 (en) * | 2018-05-31 | 2019-12-05 | The Board Of Trustees Of The Leland Stanford Junior University | MRI reconstruction using deep learning, generative adversarial network and acquisition signal model |
CN109509149A (en) * | 2018-10-15 | 2019-03-22 | 天津大学 | A kind of super resolution ratio reconstruction method based on binary channels convolutional network Fusion Features |
WO2020093042A1 (en) * | 2018-11-02 | 2020-05-07 | Deep Lens, Inc. | Neural networks for biomedical image analysis |
CN110570353A (en) * | 2019-08-27 | 2019-12-13 | 天津大学 | Dense connection generation countermeasure network single image super-resolution reconstruction method |
CN110956197A (en) * | 2019-10-28 | 2020-04-03 | 新绎健康科技有限公司 | Method and system for establishing pulse wave noise signal identification model based on convolutional neural network |
CN111723685A (en) * | 2020-06-01 | 2020-09-29 | 齐齐哈尔大学 | Remote sensing scene classification method based on branch feature fusion convolutional network |
Non-Patent Citations (2)
Title |
---|
江航;董兰芳;: "CPU环境下快速精确的人脸检测方法", 小型微型计算机***, no. 01 * |
谢朋言;: "基于深度学习的快速图像超分辨率方法", 舰船电子对抗, no. 02 * |
Also Published As
Publication number | Publication date |
---|---|
CN113762349B (en) | 2024-03-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10096134B2 (en) | Data compaction and memory bandwidth reduction for sparse neural networks | |
CN111967468B (en) | Implementation method of lightweight target detection neural network based on FPGA | |
CN113469073B (en) | SAR image ship detection method and system based on lightweight deep learning | |
CN109222972B (en) | fMRI whole brain data classification method based on deep learning | |
CN110084274B (en) | Real-time image semantic segmentation method and system, readable storage medium and terminal | |
CN111091130A (en) | Real-time image semantic segmentation method and system based on lightweight convolutional neural network | |
WO2019060670A1 (en) | Compression of sparse deep convolutional network weights | |
CN108764072A (en) | A kind of blood cell subsets image classification method based on Multiscale Fusion | |
CN113344188A (en) | Lightweight neural network model based on channel attention module | |
CN104765589B (en) | Grid parallel computation preprocess method based on MPI | |
CN116416561A (en) | Video image processing method and device | |
CN113420643B (en) | Lightweight underwater target detection method based on depth separable cavity convolution | |
CN113096096B (en) | Microscopic image bone marrow cell counting method and system fusing morphological characteristics | |
CN109726822B (en) | Operation method, device and related product | |
CN116740538A (en) | YOLOv 8-based improved lightweight target detection method and system | |
Liu et al. | EdgeYOLO: an edge-real-time object detector | |
CN114119627B (en) | High-temperature alloy microstructure image segmentation method and device based on deep learning | |
CN115115863A (en) | Water surface multi-scale target detection method, device and system and storage medium | |
CN111882053A (en) | Neural network model compression method based on splicing convolution | |
Yan et al. | Acceleration and optimization of artificial intelligence CNN image recognition based on FPGA | |
CN113344115A (en) | Target detection method based on lightweight model | |
CN113762349B (en) | Marine organism-oriented lightweight aliasing dense network classification method and system | |
CN117035047A (en) | Parallel training method and system for medical image segmentation model | |
CN111832336B (en) | Improved C3D video behavior detection method | |
US20240119269A1 (en) | Dynamic sparsity-based acceleration of neural networks |
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 |