CN106686472B - A kind of high frame-rate video generation method and system based on deep learning - Google Patents
A kind of high frame-rate video generation method and system based on deep learning Download PDFInfo
- Publication number
- CN106686472B CN106686472B CN201611241691.XA CN201611241691A CN106686472B CN 106686472 B CN106686472 B CN 106686472B CN 201611241691 A CN201611241691 A CN 201611241691A CN 106686472 B CN106686472 B CN 106686472B
- Authority
- CN
- China
- Prior art keywords
- frame
- video
- neural networks
- convolutional neural
- layer
- 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.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/80—Generation or processing of content or additional data by content creator independently of the distribution process; Content per se
- H04N21/83—Generation or processing of protective or descriptive data associated with content; Content structuring
- H04N21/845—Structuring of content, e.g. decomposing content into time segments
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/587—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal sub-sampling or interpolation, e.g. decimation or subsequent interpolation of pictures in a video sequence
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/01—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level
- H04N7/0127—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level by changing the field or frame frequency of the incoming video signal, e.g. frame rate converter
Abstract
The high frame-rate video generation method based on deep learning that the invention discloses a kind of, comprising: generate training sample set using one or more original high frame-rate video segments；The multiple video frame set training binary channels convolutional neural networks model concentrated using the training sample, with binary channels convolutional neural networks after being optimized, the binary channels convolutional neural networks model is convolutional neural networks made of being merged as two convolutional channels；Using binary channels convolutional neural networks after the optimization, the insertion frame of this two video frame is generated according to two video frame of arbitrary neighborhood in low frame-rate video, to generate the video that frame per second is higher than the low frame-rate video.The method of the present invention whole process is end to end, it does not need to carry out subsequent processing to video frame, the problems such as video frame rate conversion effect is good, and the video fluency of synthesis is high, switches for shake existing during video capture, video scene has preferable robustness.
Description
Technical field
The invention belongs to technical field of computer vision, regard more particularly, to a kind of high frame per second based on deep learning
Frequency generation method and system.
Background technique
With the development of science and technology, the mode that people obtain video is more and more convenient, however due to hardware, it is most of
Video is all that non-professional equipment is collected, and frame per second generally only has 24fps-30fps.The video of high frame per second has high smoothness
Degree, can bring better visual experience.If people directly upload the video of high frame per second on the net, due to flow
Consumption increases, and the cost of people also increases as.If the directly upper video for transmitting low frame per second, due to network line,
Inevitably there is frame losing in video, the video the big more is easy to appear this phenomenon, so that the view of distal end during transmission
Frequency quality cannot be effectively guaranteed, this greatly affected the experience of people.It is reasonable it is therefore desirable to be used in distal end
Processing mode carries out subsequent processing to the video that people upload, so that the demand that the quality of video is able to satisfy people is even further
Promote the experience of people.
Summary of the invention
Aiming at the above defects or improvement requirements of the prior art, the high frame per second based on deep learning that the present invention provides a kind of
Thus video generation method solves to regard due to low frame per second its object is to be the video of high frame per second by the Video Quality Metric of low frame per second
Frame losing of frequency during network transmission and cause video quality to decline the technical issues of affecting to the experience of people.
To achieve the above object, according to one aspect of the present invention, a kind of high frame per second view based on deep learning is provided
Frequency generation method, comprising the following steps:
(1) training sample set is generated using one or more original high frame-rate video segments, the training sample concentrates packet
It includes multiple video frame set, includes two training frames and a control frame in each video frame set, described two
Training frames are two video frames that a frame or multiframe are spaced in high frame-rate video segment, and the control frame is two training frames
Any one frame of midfeather；The frame per second of the high frame-rate video segment is higher than setting frame per second threshold value；
(2) the multiple video frame set training binary channels convolutional neural networks model concentrated using the training sample,
With binary channels convolutional neural networks after being optimized；Wherein, the binary channels convolutional neural networks model is to be led to by two convolution
Convolutional neural networks made of road fusion, two convolutional channels are respectively used to two video frames in input video frame subclass simultaneously
Convolution is carried out to the video frame of input respectively, binary channels convolutional neural networks model carries out the convolution results of two convolutional channels
It merges and exports to predict frame, the frame flyback training bilateral is compareed with the video frame set according to the prediction frame
Road convolutional neural networks model；
(3) using binary channels convolutional neural networks after the optimization, according to two video of arbitrary neighborhood in low frame-rate video
Frame generates the insertion frame of this two video frame, to generate the video that frame per second is higher than the low frame-rate video.
In one embodiment of the present of invention, each convolutional channel in the binary channels convolutional neural networks model includes k
A convolutional layer, wherein k > 0, the mathematical description of each convolutional layer are as follows:
Zi(Y)=Wi*Fi-1(Y)+Bi
Wherein i indicates that the number of plies of convolutional layer, input video frame are the 0th layer, and * represents convolution operation, Fi-1Indicate (i-1)-th layer
Output, Zi(Y) output after i-th layer of convolution operation, W are indicatediFor i-th layer of convolution nuclear parameter, BiJoin for i-th layer of biasing
Number.
In one embodiment of the present of invention, in the convolutional channel, it is connected to one respectively after preceding k-1 convolutional layer
The active coating of ReLU is to keep the sparsity of network, mathematical description are as follows:
Fi(Y)=max (0, Zi)。
In one embodiment of the present of invention, in two video frames by the feature that obtains after the last one convolutional layer
Response diagram is merged by the way of corresponding position value addition.
In one embodiment of the present of invention, swash in the Sigmoid that is followed by that the mixing operation obtains characteristic response figure
Layer living is the pixel value of picture to be mapped between 0-1, mathematical description are as follows:
In one embodiment of the present of invention, use mean value for 0, the Gaussian Profile that standard deviation is 1 initializes convolution nuclear parameter,
Biasing is initialized as 0, and benchmark learning rate is initialized as 1e-6, benchmark learning rate reduces 10 times after the m period of iteration, wherein m
For preset value.
In one embodiment of the present of invention, frame flyback instruction is compareed with the video frame set according to the prediction frame
Practice the binary channels convolutional neural networks model, specifically:
Using prediction frame with compare the error between frame, the binary channels convolution is trained using error backpropagation algorithm
Neural network；Wherein use least squares error for our majorized function, mathematical description are as follows:
Wherein i indicates i-th samples pictures, and n indicates the quantity of sample training collection, YiIndicate the video frame of neural network forecast,
Indicate the true value of corresponding video frame.
In one embodiment of the present of invention, the k value is 3；First convolutional layer has the convolution kernel of 64 9*9, step-length
For 1 pixel, Filling power 4, Filling power refers to the circle number in the zero padding of characteristic pattern periphery；Second convolutional layer has 32 1*1's
Convolution kernel, step-length are 1 pixel, Filling power 0；Third convolutional layer has the convolution kernel of 3 5*5, step-length 1, and Filling power is
2。
It is another aspect of this invention to provide that additionally providing a kind of high frame-rate video generation system based on deep learning, packet
Include training sample set generation module, binary channels convolutional neural networks optimization module and high frame-rate video generation module, in which:
The training sample set generation module, for generating training sample using one or more high frame-rate video segments
Collection, it includes two training frames in each video frame set that it includes multiple video frame set that the training sample, which is concentrated,
With a control frame, two training frames are two video frames that a frame or multiframe are spaced in high frame-rate video segment, described
Compare any one frame for the midfeather that frame is two training frames；The frame per second of the high frame-rate video segment is higher than setting frame
Rate threshold value；
The binary channels convolutional neural networks optimization module, multiple video frames for being concentrated using the training sample
Gather training binary channels convolutional neural networks model, binary channels convolutional neural networks after being optimized；Wherein, the binary channels volume
Product neural network model is the convolutional neural networks of two channels fusion, and two channels are respectively used to input the video frame
Two video frames in conjunction simultaneously carry out convolution to the video frame of input respectively, binary channels convolutional neural networks model it is logical to two
The result of road convolution is merged and is exported to predict frame, compares frame with the video frame set according to the prediction frame
Binary channels convolutional neural networks model described in regression training；
The high frame-rate video generation module is used for using binary channels convolutional neural networks after the optimization, according to low frame
Two video frame of arbitrary neighborhood in rate video generates the insertion frame of this two video frame, so that generating frame per second is higher than the low frame per second view
The video of frequency.
In one embodiment of the present of invention, each convolutional channel in the binary channels convolutional neural networks model includes k
A convolutional layer, wherein k > 0, the mathematical description of each convolutional layer are as follows:
Zi(Y)=Wi*Fi-1(Y)+Bi
Wherein i indicates that the number of plies of convolutional layer, input video frame are the 0th layer, and * represents convolution operation, Fi-1Indicate (i-1)-th layer
Output, Zi(Y) output after i-th layer of convolution operation, W are indicatediFor i-th layer of convolution nuclear parameter, BiJoin for i-th layer of biasing
Number.
In general, contemplated above technical scheme through the invention, compared with prior art, the present invention has following
Technical effect:
(1) feature extraction of the invention and the prediction of frame are obtained by the supervised learning of training sample, without artificial
Intervene, spatial diversity information can be preferably fitted under the scene of large-scale data；
(2) whole process of the invention is end to end, using the ability of self-teaching of convolutional neural networks, to pass through self
The mode of study learns model parameter, it is succinct efficiently, overcome traditional technology taken time and effort when handling video frame rate conversion and
The unconspicuous feature of effect.
Detailed description of the invention
Fig. 1 is the flow chart of the method for converting video frame rate of the invention based on deep learning, wherein FiIndicate i-th layer
Output, Yt-1、Yt、Yt+1Indicate continuous three frames video frame, YtNet is indicated for calculating error, Prediction as true value
The video frame of network prediction.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.As long as in addition, technical characteristic involved in the various embodiments of the present invention described below
Not constituting a conflict with each other can be combined with each other.
Just technical term of the invention is explained and illustrated first below:
Convolutional neural networks (Convolutional Neural Network, CNN): one kind can be used for image classification, returns
The neural network of tasks such as return, its particularity is embodied in two aspects, be on the one hand its interneuronal connection is non-complete
Connection, the weight of the connection in another aspect same layer between certain neurons is shared.Network is usually by convolutional layer, pond
Change layer and full articulamentum is constituted.Convolutional layer and pond layer are responsible for extracting the hierarchy characteristic of image, and full articulamentum is responsible for extracting
Feature classified or returned.The parameter of network includes the parameter and biasing of convolution kernel and full articulamentum, and parameter can be with
By reverse conduction algorithm from the acquistion of data middle school to.
Reverse conduction algorithm (Backpropagation Algorithm, BP): being a kind of and optimal method (such as gradient
Descent method) be used in combination, for training the common methods of artificial neural network.This method damages weight calculations all in network
The gradient of function is lost, this gradient can feed back to optimal method, for updating weight to minimize loss function.Algorithm master
To include two stages: the forward direction of excitation, backpropagation and weight update.
With the arrival of big data era, the scale of video database is also increasing, and the solution of this problem is also more next
It is more urgent.Deep neural network can by it is a kind of it is preferable in a manner of simulate the working method of human brain data analyzed,
In recent years, deep learning all achieves successful application in the every field of computer vision, but video frame rate is turned
The problem of changing there is no apparent research, and complicated in view of traditional method for converting video frame rate process, time human cost is higher, this hair
The bright one kind that proposes is based on deep learning method for converting video frame rate.This method whole process be end to end, it is easy and efficiently,
For video shake, scene switching the problems such as all there is stronger robustness.
As shown in Figure 1, may comprise steps of the present invention is based on the method for converting video frame rate of deep learning:
(1) training sample set is generated using one or more original high frame-rate video segments, the training sample concentrates packet
It includes multiple video frame set, includes two training frames and a control frame in each video frame set, described two
Training frames are two video frames that a frame or multiframe are spaced in high frame-rate video segment, and the control frame is two training frames
Any one frame of midfeather；The frame per second of the high frame-rate video segment is higher than setting frame per second threshold value；
Specifically, high frame-rate video segment can be extracted and obtain sets of video frames, obtain training sample according to a certain percentage
Collection；
Training sample set is combined by multiple video frames, and two instructions are included in each video frame set
Practice frame and a control frame.Control frame is chosen for the most intermediate of two training frames or close to that most intermediate frame.Generally
In the case of refer to and take continuous 3 frame, an intermediate frame is control frame, and another two frame is training frames；If frame per second is sufficiently high, can also take
Be separated by two frames of multiframe (depending on frame per second, cannot too many) as training frames, and interphase every multiframe in can choose middle ware
Every any one frame be control frame；Such as trained high video frame rate be 60, which has N frame, then according to interval one
The mode of this training of frame sample, from the 2nd to N-1 frame in take a frame as true value (control frame) at random, and it is the frame is adjacent
Two frames be input to inside network as training sample (two training frames).Similarly, can also come in the way of being spaced multiframe
Training sample, can be used for the video of lower frame per second in this way, i.e., the Video Quality Metric of lower frame per second is the video of high frame per second.
(2) the multiple video frame set training binary channels convolutional neural networks model concentrated using the training sample,
With binary channels convolutional neural networks after being optimized；Wherein, the binary channels convolutional neural networks model is to be led to by two convolution
Convolutional neural networks made of road fusion, two convolutional channels are respectively used to two video frames in input video frame subclass simultaneously
Convolution is carried out to the video frame of input respectively, binary channels convolutional neural networks model carries out the convolution results of two convolutional channels
It merges and exports to predict frame, the frame flyback training bilateral is compareed with the video frame set according to the prediction frame
Road convolutional neural networks model；
It first has to design and Implement a binary channels convolutional neural networks, specifically:
The binary channels convolutional neural networks model established is the convolutional neural networks of two convolutional channels fusion, includes altogether
K convolutional layer, k > 0, preferably 3 individually carry out convolution to two video frame pictures (training frames) respectively.First convolutional layer has
The convolution kernel of 64 9*9, step-length are 1 pixel, and Filling power 4, Filling power refers to the circle number in the zero padding of characteristic pattern periphery.Second
A convolutional layer has the convolution kernel of 32 1*1, and step-length is 1 pixel, Filling power 0.Third volume layer has the convolution kernel of 3 5*5,
Step-length is 1, Filling power 2.The mathematical description of convolutional layer are as follows:
Zi(Y)=Wi*Fi-1(Y)+Bi
Wherein i indicates the number of plies of network, and input picture is the 0th layer, and * represents convolution operation, Fi-1Indicate (i-1)-th layer defeated
Out, Zi(Y) output after i-th layer of convolution operation, W are indicatediFor i-th layer of convolution nuclear parameter, BiFor i-th layer of offset parameter；
In 3 convolutional layers, it is connected to the active coating of a ReLU respectively after the 1st and the 2nd convolutional layer to keep
The sparsity of network, mathematical description are as follows:
Fi(Y)=max (0, Zi)。
Two video frame pictures are added by the characteristic response figure obtained after third convolutional layer using corresponding position value
Mode merged；
After the mixing operation, obtained characteristic response figure is followed by a Sigmoid active coating with by the picture of picture
Plain value is mapped between 0-1, mathematical description are as follows:
Before the training binary channels convolutional neural networks, need to each pixel value in video frame divided by 255 into
Row normalized, the pixel value after normalization is between 0 to 1；
Also, before the training binary channels convolutional neural networks, need to initialize the use of convolutional neural networks parameter
Mean value is 0, and the Gaussian Profile that standard deviation is 1 initializes convolution nuclear parameter, and biasing is initialized as 0, the initialization of benchmark learning rate
For 1e-6, benchmark learning rate reduces 10 times after the m period of iteration, and wherein m is preset value；For example, m preferably 2, then in preceding 1-m
In a iteration cycle, learning rate=1e-6, after the m period of iteration, learning rate=1e-7, and be always maintained at constant.
Specifically, can use the predicted value of network with compare between error, instructed using error backpropagation algorithm
Practice binary channels convolutional neural networks.Use least squares error for our majorized function, mathematical description are as follows:
Wherein i indicates i-th samples pictures, and n indicates the quantity of sample training collection, YiIndicate the video frame of neural network forecast,
Indicate the true value of corresponding video frame；
(3) using binary channels convolutional neural networks after the optimization, according to two video of arbitrary neighborhood in low frame-rate video
Frame generates the insertion frame of this two video frame, to generate the video that frame per second is higher than the low frame-rate video.
As it will be easily appreciated by one skilled in the art that the foregoing is merely illustrative of the preferred embodiments of the present invention, not to
The limitation present invention, any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should all include
Within protection scope of the present invention.
Claims (3)
1. a kind of high frame-rate video generation method based on deep learning, which is characterized in that the described method comprises the following steps:
(1) training sample set is generated using one or more original high frame-rate video segments, it includes more that the training sample, which is concentrated,
A video frame set includes two training frames and a control frame, two training in each video frame set
Frame is two video frames that a frame or multiframe are spaced in high frame-rate video segment, and the control frame is among two training frames
Any one frame at interval；The frame per second of the high frame-rate video segment is higher than setting frame per second threshold value；
(2) the multiple video frame set training binary channels convolutional neural networks model concentrated using the training sample, to obtain
Binary channels convolutional neural networks after must optimizing；Wherein, the binary channels convolutional neural networks model is to be melted by two convolutional channels
Convolutional neural networks made of conjunction, two convolutional channels are respectively used to two video frames and difference in input video frame subclass
Convolution is carried out to the video frame of input, binary channels convolutional neural networks model merges the convolution results of two convolutional channels
And export to predict frame, it is rolled up according to the prediction frame with the frame flyback training binary channels that compares in the video frame set
Product neural network model；Wherein,
Each convolutional channel in the binary channels convolutional neural networks model includes k convolutional layer, wherein k > 0, each convolution
The mathematical description of layer are as follows:
Zi(Y)=Wi*Fi-1(Y)+Bi
Wherein i indicates that the number of plies of convolutional layer, input video frame are the 0th layer, and * represents convolution operation, Fi-1Indicate (i-1)-th layer defeated
Out, Zi(Y) output after i-th layer of convolution operation, W are indicatediFor i-th layer of convolution nuclear parameter, BiFor i-th layer of offset parameter；
In the convolutional channel, it is connected to the active coating of a ReLU respectively after preceding k-1 convolutional layer to keep network
Sparsity, mathematical description are as follows:
Fi(Y)=max (0, Zi)；
Use mean value for 0, the Gaussian Profile that standard deviation is 1 initializes convolution nuclear parameter, and biasing is initialized as 0, benchmark study speed
Rate is initialized as 1e-6, benchmark learning rate reduces 10 times after the m period of iteration, and wherein m is preset value；
The k value is 3；First convolutional layer has the convolution kernel of 64 9*9, and step-length is 1 pixel, Filling power 4, Filling power
Refer to the circle number in the zero padding of characteristic pattern periphery；Second convolutional layer has the convolution kernel of 32 1*1, and step-length is 1 pixel, Filling power
It is 0；Third convolutional layer has the convolution kernel of 3 5*5, step-length 1, Filling power 2；
A Sigmoid active coating is followed by what the mixing operation obtained characteristic response figure the pixel value of picture to be mapped to
Between 0-1, mathematical description are as follows:
The frame flyback training binary channels convolutional neural networks are compareed with the video frame set according to the prediction frame
Model, specifically:
Using prediction frame with compare the error between frame, the binary channels convolutional Neural is trained using error backpropagation algorithm
Network；Wherein use least squares error for majorized function, mathematical description are as follows:
Wherein i indicates i-th samples pictures, and n indicates the quantity of sample training collection, YiIndicate the video frame of neural network forecast,It indicates
The true value of corresponding video frame；
(3) raw according to two video frame of arbitrary neighborhood in low frame-rate video using binary channels convolutional neural networks after the optimization
At the insertion frame of this two video frame, to generate the video that frame per second is higher than the low frame-rate video.
2. the high frame-rate video generation method based on deep learning as described in claim 1, which is characterized in that at described two
Video frame is merged by the way of corresponding position value addition by the characteristic response figure obtained after the last one convolutional layer.
3. a kind of high frame-rate video based on deep learning generates system, which is characterized in that including training sample set generation module,
Binary channels convolutional neural networks optimization module and high frame-rate video generation module, in which:
The training sample set generation module, for generating training sample set, institute using one or more high frame-rate video segments
Stating training sample and concentrating includes multiple video frame set, includes two training frames and one in each video frame set
Frame is compareed, two training frames are two video frames that a frame or multiframe are spaced in high frame-rate video segment, the control frame
For any one frame of the midfeather of two training frames；The frame per second of the high frame-rate video segment is higher than setting frame per second threshold
Value；
The binary channels convolutional neural networks optimization module, multiple video frame set for being concentrated using the training sample
Training binary channels convolutional neural networks model, binary channels convolutional neural networks after being optimized；Wherein, the binary channels convolution mind
It is the convolutional neural networks of two channels fusion through network model, two channels are respectively used to input in the video frame set
Two video frames and convolution carried out respectively to the video frame of input, binary channels convolutional neural networks model rolls up two channels
Long-pending result is merged and is exported to predict frame, compares frame flyback with the video frame set according to the prediction frame
The training binary channels convolutional neural networks model；
The high frame-rate video generation module, for being regarded according to low frame per second using binary channels convolutional neural networks after the optimization
Two video frame of arbitrary neighborhood in frequency generates the insertion frame of this two video frame, to generate frame per second higher than the low frame-rate video
Video；
Each convolutional channel in the binary channels convolutional neural networks model includes k convolutional layer, wherein k > 0, each convolution
The mathematical description of layer are as follows:
Zi(Y)=Wi*Fi-1(Y)+Bi
Wherein i indicates that the number of plies of convolutional layer, input video frame are the 0th layer, and * represents convolution operation, Fi-1Indicate (i-1)-th layer defeated
Out, Zi(Y) output after i-th layer of convolution operation, W are indicatediFor i-th layer of convolution nuclear parameter, BiFor i-th layer of offset parameter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611241691.XA CN106686472B (en) | 2016-12-29 | 2016-12-29 | A kind of high frame-rate video generation method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611241691.XA CN106686472B (en) | 2016-12-29 | 2016-12-29 | A kind of high frame-rate video generation method and system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106686472A CN106686472A (en) | 2017-05-17 |
| CN106686472B true CN106686472B (en) | 2019-04-26 |
Family
ID=58872327
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611241691.XA Active CN106686472B (en) | 2016-12-29 | 2016-12-29 | A kind of high frame-rate video generation method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106686472B (en) |
Families Citing this family (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9756375B2 (en) | 2015-01-22 | 2017-09-05 | Microsoft Technology Licensing, Llc | Predictive server-side rendering of scenes |
| CN107481209B (en) * | 2017-08-21 | 2020-04-21 | 北京航空航天大学 | Image or video quality enhancement method based on convolutional neural network |
| CN107613299A (en) * | 2017-09-29 | 2018-01-19 | 杭州电子科技大学 | A kind of method for improving conversion effect in frame rate using network is generated |
| CN107886081B (en) * | 2017-11-23 | 2021-02-02 | 武汉理工大学 | Intelligent underground dangerous behavior grading identification method for double-path U-Net deep neural network mine |
| CN108111860B (en) * | 2018-01-11 | 2020-04-14 | 安徽优思天成智能科技有限公司 | Video sequence lost frame prediction recovery method based on depth residual error network |
| CN108322685B (en) * | 2018-01-12 | 2020-09-25 | 广州华多网络科技有限公司 | Video frame insertion method, storage medium and terminal |
| CN108600655A (en) * | 2018-04-12 | 2018-09-28 | 视缘(上海)智能科技有限公司 | A kind of video image synthetic method and device |
| CN108600762B (en) * | 2018-04-23 | 2020-05-15 | 中国科学技术大学 | Progressive video frame generation method combining motion compensation and neural network algorithm |
| CN108830812B (en) * | 2018-06-12 | 2021-08-31 | 福建帝视信息科技有限公司 | Video high frame rate reproduction method based on grid structure deep learning |
| CN108810551B (en) * | 2018-06-20 | 2021-01-12 | Oppo（重庆）智能科技有限公司 | Video frame prediction method, terminal and computer storage medium |
| CN108961236B (en) * | 2018-06-29 | 2021-02-26 | 国信优易数据股份有限公司 | Circuit board defect detection method and device |
| CN110780664A (en) * | 2018-07-25 | 2020-02-11 | 格力电器（武汉）有限公司 | Robot control method and device and sweeping robot |
| CN109379550B (en) * | 2018-09-12 | 2020-04-17 | 上海交通大学 | Convolutional neural network-based video frame rate up-conversion method and system |
| CN109068174B (en) * | 2018-09-12 | 2019-12-27 | 上海交通大学 | Video frame rate up-conversion method and system based on cyclic convolution neural network |
| CN109120935A (en) * | 2018-09-27 | 2019-01-01 | 贺禄元 | A kind of coding method of video image and device |
| US10924525B2 (en) | 2018-10-01 | 2021-02-16 | Microsoft Technology Licensing, Llc | Inducing higher input latency in multiplayer programs |
| CN109360436B (en) * | 2018-11-02 | 2021-01-08 | Oppo广东移动通信有限公司 | Video generation method, terminal and storage medium |
| CN110163061B (en) * | 2018-11-14 | 2023-04-07 | 腾讯科技（深圳）有限公司 | Method, apparatus, device and computer readable medium for extracting video fingerprint |
| CN111371983A (en) * | 2018-12-26 | 2020-07-03 | 清华大学 | Video online stabilization method and system |
| CN113766313B (en) * | 2019-02-26 | 2024-03-05 | 深圳市商汤科技有限公司 | Video data processing method and device, electronic equipment and storage medium |
| JP7201073B2 (en) * | 2019-04-01 | 2023-01-10 | 株式会社デンソー | Information processing equipment |
| CN110636221A (en) * | 2019-09-23 | 2019-12-31 | 天津天地人和企业管理咨询有限公司 | System and method for super frame rate of sensor based on FPGA |
| CN112584158B (en) * | 2019-09-30 | 2021-10-15 | 复旦大学 | Video quality enhancement method and system |
| WO2021104381A1 (en) * | 2019-11-27 | 2021-06-03 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method and device for stylizing video and storage medium |
| CN113630621B (en) * | 2020-05-08 | 2022-07-19 | 腾讯科技（深圳）有限公司 | Video processing method, related device and storage medium |
| RU2747965C1 (en) * | 2020-10-05 | 2021-05-18 | Самсунг Электроникс Ко., Лтд. | Frc occlusion processing with deep learning |
| US11889227B2 (en) | 2020-10-05 | 2024-01-30 | Samsung Electronics Co., Ltd. | Occlusion processing for frame rate conversion using deep learning |
| CN113516050A (en) * | 2021-05-19 | 2021-10-19 | 江苏奥易克斯汽车电子科技股份有限公司 | Scene change detection method and device based on deep learning |
| CN113420771B (en) * | 2021-06-30 | 2024-04-19 | 扬州明晟新能源科技有限公司 | Colored glass detection method based on feature fusion |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202285412U (en) * | 2011-09-02 | 2012-06-27 | 深圳市华美特科技有限公司 | Low frame rate transmission or motion image twinkling elimination system |
| CN104102919A (en) * | 2014-07-14 | 2014-10-15 | 同济大学 | Image classification method capable of effectively preventing convolutional neural network from being overfit |
| CN105787510A (en) * | 2016-02-26 | 2016-07-20 | 华东理工大学 | System and method for realizing subway scene classification based on deep learning |
| CN106022237A (en) * | 2016-05-13 | 2016-10-12 | 电子科技大学 | Pedestrian detection method based on end-to-end convolutional neural network |
| CN106228124A (en) * | 2016-07-17 | 2016-12-14 | 西安电子科技大学 | SAR image object detection method based on convolutional neural networks |
-
2016
- 2016-12-29 CN CN201611241691.XA patent/CN106686472B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202285412U (en) * | 2011-09-02 | 2012-06-27 | 深圳市华美特科技有限公司 | Low frame rate transmission or motion image twinkling elimination system |
| CN104102919A (en) * | 2014-07-14 | 2014-10-15 | 同济大学 | Image classification method capable of effectively preventing convolutional neural network from being overfit |
| CN105787510A (en) * | 2016-02-26 | 2016-07-20 | 华东理工大学 | System and method for realizing subway scene classification based on deep learning |
| CN106022237A (en) * | 2016-05-13 | 2016-10-12 | 电子科技大学 | Pedestrian detection method based on end-to-end convolutional neural network |
| CN106228124A (en) * | 2016-07-17 | 2016-12-14 | 西安电子科技大学 | SAR image object detection method based on convolutional neural networks |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106686472A (en) | 2017-05-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106686472B (en) | A kind of high frame-rate video generation method and system based on deep learning | |
| CN109064507B (en) | Multi-motion-stream deep convolution network model method for video prediction | |
| Zhou et al. | Photorealistic facial expression synthesis by the conditional difference adversarial autoencoder | |
| Feng et al. | SGANVO: Unsupervised deep visual odometry and depth estimation with stacked generative adversarial networks | |
| CN112669325B (en) | Video semantic segmentation method based on active learning | |
| CN108805083A (en) | The video behavior detection method of single phase | |
| CN105095862B (en) | A kind of human motion recognition method based on depth convolution condition random field | |
| WO2022252272A1 (en) | Transfer learning-based method for improved vgg16 network pig identity recognition | |
| CN108830913B (en) | Semantic level line draft coloring method based on user color guidance | |
| CN111489372B (en) | Video foreground and background separation method based on cascade convolution neural network | |
| CN111444878A (en) | Video classification method and device and computer readable storage medium | |
| CN107590432A (en) | A kind of gesture identification method based on circulating three-dimensional convolutional neural networks | |
| CN106952271A (en) | A kind of image partition method handled based on super-pixel segmentation and EM/MPM | |
| CN110443784B (en) | Effective significance prediction model method | |
| CN108647599B (en) | Human behavior recognition method combining 3D (three-dimensional) jump layer connection and recurrent neural network | |
| CN111241963B (en) | First person view video interactive behavior identification method based on interactive modeling | |
| CN113807318B (en) | Action recognition method based on double-flow convolutional neural network and bidirectional GRU | |
| CN109345446A (en) | A kind of image style branching algorithm based on paired-associate learning | |
| Simonyan et al. | Two-stream convolutional networks for action recognition | |
| CN105976379A (en) | Fuzzy clustering color image segmentation method based on cuckoo optimization | |
| Tan et al. | Bidirectional long short-term memory with temporal dense sampling for human action recognition | |
| CN109583334A (en) | A kind of action identification method and its system based on space time correlation neural network | |
| Desai et al. | Next frame prediction using ConvLSTM | |
| CN116012950A (en) | Skeleton action recognition method based on multi-heart space-time attention pattern convolution network | |
| CN112365428B (en) | DQN-based highway monitoring video defogging method and system |
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 |