CN106845351A - It is a kind of for Activity recognition method of the video based on two-way length mnemon in short-term - Google Patents
It is a kind of for Activity recognition method of the video based on two-way length mnemon in short-term Download PDFInfo
- Publication number
- CN106845351A CN106845351A CN201611193290.1A CN201611193290A CN106845351A CN 106845351 A CN106845351 A CN 106845351A CN 201611193290 A CN201611193290 A CN 201611193290A CN 106845351 A CN106845351 A CN 106845351A
- Authority
- CN
- China
- Prior art keywords
- video
- frame
- short
- feature
- term
- 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.)
- Pending
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
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- 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
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- 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
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
Abstract
The invention discloses a kind of for Activity recognition method of the video based on two-way length mnemon in short-term, including:(1) input video sequence, extracts the RBG frame sequences and light stream picture in video sequence;(2) RGB image depth convolutional network and light stream picture depth convolutional network are respectively trained;(3) multilayer feature of network is extracted, wherein at least extracts the 3rd convolutional layer, the 5th convolutional layer, the feature of the 7th full articulamentum;He Chihua is carried out to convolutional layer feature;(4) to being trained using the two-way length recurrent neural network that mnemon builds in short-term, probability matrix of the video per frame is obtained;(5) each probability matrix is averaged, finally merges the probability matrix of light stream frame and RGB frame, take the class of maximum probability as last classification results, be achieved in Activity recognition.The present invention replaces traditional manual features, the depth characteristic of different layers to characterize different information using the feature of multilayer deep learning, and the combination of multilayer feature can improve the accuracy rate of classification;By using two-way short-term memory capture time information long, more spatial structure information are obtained, improve the effect of Activity recognition.
Description
Technical field
The present invention relates to a kind of method for processing video frequency, and in particular to the side of the behavior of personage in a kind of automatic identification video
Method.
Background technology
Activity recognition refer to by extracting video or image sequence in behavior of the characteristic information to target be analyzed, with
Personage's behavior pattern in identification video.
Activity recognition is computer vision and an important and difficult problem in pattern-recognition.It is in many aspects
Have a wide range of applications, such as intelligent monitoring, man-machine interaction, virtual reality, intelligent security guard etc..Today's society is with economy
Fast development, safety problem increasingly paid close attention to by people, and increasing place is assembled with video monitoring camera,
There is substantial amounts of monitor video to produce daily, present people generally take and are manually monitored, and this is accomplished by putting into substantial amounts of people
Power, material resources, financial resources;In addition, people in prolonged video monitoring, notice can decline, cause some emergencies without
Method is arrived by effective detection, has delayed the time of the work such as rescue, causes substantial amounts of loss.In order to solve this problem, if can
Monitored with by computer automatic video frequency, then will effectively reduce the input of this respect, and timely warning information can be obtained,
Taken measures with facilitating.
The conventional method of Activity recognition is to extract the feature of frame of video first, is then identified using disaggregated model.Its
In, extract that effective feature representation is critically important from video, directly affect the performance of whole system.
Traditional Activity recognition method, uses different hand-designed features.Deep learning method then utilizes neutral net
The feature for learning, non-linear due to depth network, it can include more internal informations, can be with significant increase mark sheet
Up to generalization ability.
Traditional behavioural characteristic extracting method mainly has:
1)Method based on low-level feature such as moving object detection, tracking.The method is to extract low layer pictures feature to be analyzed,
Method is fairly simple directly perceived.The low-level feature that can be used mainly includes movement velocity, direction, light stream, the target of foreground target
Shape contour, movement locus etc..This category feature can ignore static irrelevant information, pay close attention to moving target, it is possible to reduce
The interference of irrelevant information in frame of video.In addition, the method to extract feature also relatively easy, but wherein there is also some and serious ask
Topic.Such as, this category feature heavy dependence target tracking algorism, if tracking effect is too poor, it will cause pole to last result
Big influence.And the video in real world, various interference, such as mixed and disorderly background, other moving targets are often included again.This
This kind of method is resulted in carry out daily video behavior analytical effect poor, thus this category feature often robustness is poor, it is difficult to
It is applied in actual scene.
2)The method that subcharacter is described based on space-time.The behavior pattern of this kind of characteristic use craft feature description.When such as
Empty interest point methods, it is described with some points not associated to human action.In addition, also intensive track (Dense
Trajectories), track (Improved Trajectories) is improved, Scale invariant features transform (SIFT), gradient Nogata
Figure (HOG), the method such as light stream histogram (HOF), the simple motion feature more robust of the aspect ratio of these manual designs, but it is local
The amount of calculation for describing son is larger, is also easily disturbed by noise.Current these features description achieves good effect, but its
The space of lifting very little, it is necessary to find more powerful effective feature.
3)Feature based on middle level semantic understanding.This category information generally uses a unified manikin, this model
Human body is divided into different parts, such as head, shoulder, arm, leg, this expression can obtain precision higher, also compare
Compared with robust, but the structure of model is relative complex, it is necessary to substantial amounts of work.In order to lift accuracy rate, many articles utilize RGB-D depths
Degree figure is detected.
Three of the above is the main feature used in traditional Activity recognition.The wise and able performance in order to lift Activity recognition of woods,
Effective integration is carried out using multiple features, certain achievement is achieved.Lai et al. per frame as an example, using how real
Example learning method, system can be while reasoning frame tagging and video tab.Haoi et al. goes to model frame of video using structuring SVM
Between relation, and predict whole event in observation portion event.
Conventional learning algorithmses achieve good result to a certain extent.But with the appearance of deep learning, many necks
The performance in domain is obtained for greatly lifting.Deep learning simulates the working mechanism of brain, and it employs the network structure of multilayer,
It is a kind of nonlinear model, abstract by carrying out to visual object with powerful data capability of fitting and learning ability,
Can be with unsupervised from data learning to internal information.The feature obtained by deep learning, the sensing results phase with the mankind
Seemingly, certain semantic information has been usually contained, identification has more conducively been analyzed to visual object.
Static images classification is carried out using convolutional neural networks have been achieved with greatly success.But because the complexity of video
Structure and noise jamming, the attention rate of Activity recognition are less.If we are processed every frame frame of video as static images,
Then the result of all frames of video is averaging and draws and classify belonging to this video.But such method relies only on static frames loses
Too many time related information is lost, can so cause hydraulic performance decline.
For Activity recognition, two problems are primarily present.
1st, the wide gap between feature and video semanteme is extracted.The feature of mankind's hand-designed, achieved effectively in the past
Achievement, but present manual feature representation is to a bottleneck, it is difficult to obtain big lifting again, particularly processes video
The problems such as this challenge, the ambient interferences of video, frame per second change, illumination variation, motion of visual angle change and video camera etc.
The performance of system is all had a strong impact on, so needing a more preferable feature description.
2nd, another problem for meriting attention is how to model the temporal information of video.Using only space characteristics, it is impossible to
Accurate description video.In order to solve this problem, it is necessary to consider to add temporal information.Different actions possess different sequences,
Time-domain information is added can effectively lift recognition effect.
The content of the invention
Goal of the invention of the invention is to provide a kind of for Activity recognition of the video based on two-way length mnemon in short-term
Method, traditional manual features are replaced by the feature using multilayer deep learning, the accuracy rate of classification are improved, by using double
To short-term memory long (Bi-directional Long Short-Term Memory) capture time information, when obtaining more
Domain structure information, to improve the effect of Activity recognition.
To achieve the above object of the invention, the technical solution adopted by the present invention is:It is a kind of for video based on two-way length
When mnemon Activity recognition method, comprise the following steps:
(1) input video sequence, extracts the RBG frame sequences and light stream picture in the video sequence;
(2) depth convolutional network is trained:It is respectively trained RGB image depth convolutional network and light stream picture depth convolutional network;
(3) multilayer feature of network is extracted, wherein at least extracts the 3rd convolutional layer, the 5th convolutional layer, the spy of the 7th full articulamentum
Levy;A vector for fixed size is obtained to full articulamentum;He Chihua is carried out to convolutional layer feature, temporal information is added;
(4) using each layer characteristic vector obtained from convolutional neural networks, to using two-way length, mnemon structure is passed in short-term
Neutral net is returned to be trained, and input test collection obtains the probability matrix and RGB of the every frame of light stream frame in video using Softmax
Probability matrix of the frame per frame;
(5) probability matrix respectively to all light stream frames in video is averaged, and the probability matrix to all RGB frames in video takes
Averagely, the probability matrix of light stream frame and RGB frame is finally merged, the class of maximum probability is taken as last classification results, it is thus real
Existing Activity recognition.
Preferred technical scheme, in step (2), the 3rd convolutional layer of selection, the 5th convolutional layer and the 7th full articulamentum conduct
Feature representation.
Described and pond turns to:
It is in m layers of characteristic pattern of time t, pond is carried out using following formula:
Wherein, j is pond frame, and N is time domain pond scope, finally describes the dimension of son and is,It is feature
The height of figure,It is the width of characteristic pattern, C is characteristic pattern channel number, and T is the totalframes of video.
Further technical scheme, before step (4) behind described and pond, using spatial pyramid maximum pond, makes
Two-layer pyramid is used, the characteristic vector that dimension is fixed length is obtained.
Preferentially, N is 15.
Because above-mentioned technical proposal is used, the present invention has following advantages compared with prior art:
1st, the present invention replaces traditional manual features, the depth characteristic of different layers to characterize not using the feature of multilayer deep learning
Same information, the combination of multilayer feature can improve the accuracy rate of classification.
2nd, the present invention is using two-way short-term memory (Bi-directional Long Short-Term Memory) capture long
Temporal information.Two-way short-term memory long is usually used to treatment sequence problem, and mnemon can only be to folk prescription in short-term for unidirectional length
It is modeled to the time, but present frame will not only refer to former frame sometimes, it is also contemplated that the information of a later frame, so, using double
Just both direction can be modeled to mnemon in short-term long, it is ensured that the accuracy of Activity recognition.
Brief description of the drawings
Fig. 1 is the method frame composition of the embodiment of the present invention;
Fig. 2 is the visualization of the characteristic pattern to identical input picture different layers;
Fig. 3 is LSTM configuration diagrams;
Fig. 4 is the two-way LSTM configuration diagrams in the embodiment of the present invention;
Fig. 5 is UCF101 data set schematic diagrames;
Fig. 6 is the visualization schematic diagram of different layers.
Specific embodiment
Below in conjunction with the accompanying drawings and embodiment the invention will be further described:
Embodiment one:It is shown in Figure 1, it is a kind of for Activity recognition side of the video based on two-way length mnemon in short-term
Method, using deep learning to feature with reference to two-way length, mnemon carries out Activity recognition in short-term.In order to select one it is strong
Feature representation, use the traditional hand-designed feature of the deep learning feature replacement of multilayer, improve Activity recognition performance.For
Abundant excavation timing information, using two-way length, mnemon (Bi-LSTM) is modeled in short-term, it can capture both direction
The change of upper time series, there is provided information be better than unidirectional length mnemon in short-term.
1st, convolutional neural networks and its multilayer feature
In order to extract effectively expressing feature, it is necessary to train a depth convolutional network.The present embodiment uses simple and effective
Caffe frameworks build system.Pre-training is carried out using ImageNet data the set pair analysis model.ImageNet data sets are comprising substantial amounts of
Picture, this can ensure the generalization ability of model.Afterwards, the good model of pre-training is migrated to be finely adjusted to real data collection.
Use the network architecture of similar double fluid.The light stream of RGB frame and video sequence is extracted in advance and is saved in local disk.The present embodiment
Using the frame of 224 × 224 sizes as input.
Different layers include different information.The preceding several layers of of network have more low-level features, such as marginal information.Network
It is several layers of afterwards can be more abstract, encode the more semantic informations of video.Fig. 2 is the characteristic pattern that picture different layers are input into identical
Visualization.From figure, it has been found that different convolutional layers are represented comprising different information.They have different to same width picture
Response.We are by experiment the 3rd convolutional layer (conv3) of last selection, the 5th convolutional layer (conv5) and the 7th full articulamentum
(fc7) as our feature representation.The generalization ability of the first convolutional layer (conv1) and the second convolutional layer (conv2) is too poor, right
Last recognition result effect is little.
The parameter of each layer of convolutional neural networks is as follows:
After characteristic pattern is extracted, characteristic pattern is used and pond (Sum Pooling).If it is in m layers of characteristic pattern of time t,
Pond is carried out using following formula:
Wherein, j is pond frame, and N is time domain pond scope.Here 15 frames are selected.Finally describing the dimension of son is。It is the height of characteristic pattern,It is the width of characteristic pattern, C is characteristic pattern channel number, and T is the totalframes of video.For connecting entirely
Layer is connect, 4096 dimensional feature vectors for directly being exported using network.
2nd, recurrent neural network
Mainly there are two kinds of neutral net, feedforward neural network and recurrent neural network in the prior art.Feedforward Neural Networks
Network has been achieved for successfully in numerous applications, but it but cannot very well process sequence problem.Conversely, recurrent neural network
Feature is allowed to be modeled time domain very simple.Recurrent neural network is using sequence as input.For video sequence, output with
Present frame is relevant with former frame.If given list entries is expressed as, there is equation below:
WhereinThe activation value of the hidden layer in time t is represented,Input layer to the weight matrix of hidden layer is represented,Table
Show the weight matrix between hidden layer,It is skew, f is activation primitive.Finally, exported by equation below:
Wherein,Hidden layer to the weight matrix of output layer is represented,It is output offset.
RNN subject matters are that it can only effectively be modeled to the sequence of short time, as network depth is deepened, can be caused
Gradient disperse, frame in the past is too small to current action effect.In order to solve the problem, mnemon (LSTM) in short-term long is introduced
Three doors go to keep network state.Three doors are respectively input gates, forget door and out gate.By adding before input gate control
These three doors, it makes in sequence problem is processed, and graded is more steady.LSTM is using one than conventional recursive neutral net
Complicated framework goes to improve the performance for the treatment of time series long.Fig. 3 describes LSTM frameworks.
3rd, two-way length mnemon in short-term
Although LSTM can capture sequential column information long, it is unidirectional.That is, before present frame is only received in LSTM
Frame influences.In the present invention, in order to strengthen this relation, it is allowed to be extended to two-way.That is, treatment when the current frame, it is also contemplated that
The influence of frame afterwards.Its implementation is, processes from back to front along with one layer.Two-way LSTM models such as Fig. 4 of the present embodiment
It is shown.Ground floor is to process sequence from front to back, and the second layer is to process sequence from back to front.Finally, the result of this two-layer can be total to
Same-action is input in Softmax graders.Unidirectional LSTM can only capture the time-evolution information in direction, but two-way LSTM
Two-way sequential organization can be modeled, therefore it can capture more spatial structure information.
The model is tested on UCF101 and HMDB51 data sets.UCF101 data sets include 13,320 video sequences,
Totally 101 behavior classifications.As shown in figure 5, these videos are shot by professional.It is divided into three parts, each portion
It is respectively divided training set and test set.The method of the present embodiment is tested on each part and by results averaged.
HMDB51 data sets come from many aspects, such as film or Internet video.This data set is made up of 6766 video sequences,
Totally 51 behavior classifications.It has more challenge than other data sets, because its background environment is more complicated.Data set is also divided
It is three parts.Three averages of partial results are taken as last result.
In order to choose useful CNN layers, the experimental result of different layers is compared on UCF101 data sets.Can be with from Fig. 6
It was found that, for width picture input, different layers have different activation.It is preceding it is several layers of have more detailed information, it is rear several layers of more to take out
As.But it has also been found that the first convolutional layer and the second convolutional layer include excessive interference information.
Next step carries out quantitative analysis.For simple and quick comparing, each video is only sampled 10 frames.Short-sighted frequency is on the one hand special
Levy representative enough, on the other hand reduce experimental period.Result in table 1 is with RGB frame on spatial network
As the result that input is produced.From table 1, can also find that earliest several layers of effects are not best, this can influence last result.
The different individual layer performances on UCF101 of table 1
Layer | conv1 | conv2 | conv3 | conv4 | conv5 | fc6 | fc7 |
Accuracy rate | 33.5% | 44.7% | 68.2% | 68.9% | 69.1% | 60.3% | 61.4% |
Continue the result of different layers combination.Result more more preferable than individual layer can be obtained using multiple layer combination.This demonstrate that multilayer
CNN features are more powerful.Such as table 2, find to use the 3rd convolutional layer (conv3), the 5th convolutional layer (conv5) and the in an experiment
Seven full articulamentums (fc7) obtain best result.
Performance of the various combination of table 2 layer on UCF101
The number of plies | Accuracy rate |
conv3 + conv4 | 70.1% |
conv4 + conv5 | 69.2% |
conv3 + conv5 | 70.5% |
conv3 + conv5 + fc6 | 70.3% |
conv3 + conv5 + fc7 | 70.8% |
Again to the two-way length of the present embodiment in short-term mnemon (Bi-LSTM) and averaging model and unidirectional unit in short-term long carry out it is right
Than experiment.Averaging model directly go to obtain last result by the score of average each Softmax.Using individual layer LSTM in individual layer
In LSTM models.All three model has been proposed that fc7 as input.Find two-way LSTM better than average from the result in table 3
Model and unidirectional LSTM.
Different model performances on the UCF101 of table 3
Model | Accuracy rate |
Averaging model | 61.4% |
Unidirectional LSTM | 62.1% |
Two-way LSTM | 63.5% |
Finally, table 4 describes the accuracy rate of various methods.The method that proof adds multi-layer C NN features and two-way LSTM can be with
Effectively improve Activity recognition performance.
Result of the distinct methods of table 4 on data set UCF101 and HMDB51
Model | UCF101 | HMDB51 |
STIP+BovW (2011, 2012) | 43.9% | 23.0% |
Motionlets (2013) | - | 42.1% |
DT+MVSV(2014) | 83.5% | 55.9% |
iDT+FV (2013) | 85.9% | 57.2% |
iDT+HSV (2014) | 87.9% | 61.1% |
Two-Stream (2014) | 88.0% | 59.4% |
LRCN (2015) | 82.9% | - |
BSS (2015) | 88.6% | - |
Composite LSTM (2015) | 84.3% | - |
The present embodiment | 88.9% | 62.3% |
Claims (5)
1. a kind of for Activity recognition method of the video based on two-way length mnemon in short-term, comprise the following steps:
(1) input video sequence, extracts the RBG frame sequences and light stream picture in the video sequence;
(2) depth convolutional network is trained:It is respectively trained RGB image depth convolutional network and light stream picture depth convolutional network;
(3) multilayer feature of network is extracted, wherein at least extracts the 3rd convolutional layer, the 5th convolutional layer, the spy of the 7th full articulamentum
Levy;A vector for fixed size is obtained to full articulamentum;He Chihua is carried out to convolutional layer feature, temporal information is added;
(4) using each layer characteristic vector obtained from convolutional neural networks, to using two-way length, mnemon structure is passed in short-term
Neutral net is returned to be trained, and input test collection obtains the probability matrix and RGB of the every frame of light stream frame in video using Softmax
Probability matrix of the frame per frame;
(5) probability matrix respectively to all light stream frames in video is averaged, and the probability matrix to all RGB frames in video takes
Averagely, the probability matrix of light stream frame and RGB frame is finally merged, the class of maximum probability is taken as last classification results, it is thus real
Existing Activity recognition.
2. according to claim 1 for Activity recognition method of the video based on two-way length mnemon in short-term, it is special
Levy and be:In step (2), the 3rd convolutional layer of selection, the 5th convolutional layer and the 7th full articulamentum are used as feature representation.
3. according to claim 1 for Activity recognition method of the video based on two-way length mnemon in short-term, it is special
Levy and be, described and pond turns to:
It is in m layers of characteristic pattern of time t, pond is carried out using following formula:
Wherein, j is pond frame, and N is time domain pond scope, finally describes the dimension of son and is,It is feature
The height of figure,It is the width of characteristic pattern, C is characteristic pattern channel number, and T is the totalframes of video.
4. according to claim 3 for Activity recognition method of the video based on two-way length mnemon in short-term, it is special
Levy and be:Before step (4) behind described and pond, using spatial pyramid maximum pond, using two-layer pyramid, tieed up
Spend the characteristic vector for fixed length.
5. according to claim 3 for Activity recognition method of the video based on two-way length mnemon in short-term, it is special
Levy and be:N is 15.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2016103160124 | 2016-05-13 | ||
CN201610316012 | 2016-05-13 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106845351A true CN106845351A (en) | 2017-06-13 |
Family
ID=59135213
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611193290.1A Pending CN106845351A (en) | 2016-05-13 | 2016-12-21 | It is a kind of for Activity recognition method of the video based on two-way length mnemon in short-term |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106845351A (en) |
Cited By (67)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107330463A (en) * | 2017-06-29 | 2017-11-07 | 南京信息工程大学 | Model recognizing method based on CNN multiple features combinings and many nuclear sparse expressions |
CN107423721A (en) * | 2017-08-08 | 2017-12-01 | 珠海习悦信息技术有限公司 | Interactive action detection method, device, storage medium and processor |
CN107463949A (en) * | 2017-07-14 | 2017-12-12 | 北京协同创新研究院 | A kind of processing method and processing device of video actions classification |
CN107463879A (en) * | 2017-07-05 | 2017-12-12 | 成都数联铭品科技有限公司 | Human bodys' response method based on deep learning |
CN107484017A (en) * | 2017-07-25 | 2017-12-15 | 天津大学 | Supervision video abstraction generating method is had based on attention model |
CN107506712A (en) * | 2017-08-15 | 2017-12-22 | 成都考拉悠然科技有限公司 | Method for distinguishing is known in a kind of human behavior based on 3D depth convolutional networks |
CN107644519A (en) * | 2017-10-09 | 2018-01-30 | 中电科新型智慧城市研究院有限公司 | A kind of intelligent alarm method and system based on video human Activity recognition |
CN107679522A (en) * | 2017-10-31 | 2018-02-09 | 内江师范学院 | Action identification method based on multithread LSTM |
CN108009493A (en) * | 2017-11-30 | 2018-05-08 | 电子科技大学 | Face anti-fraud recognition methods based on action enhancing |
CN108038103A (en) * | 2017-12-18 | 2018-05-15 | 北京百分点信息科技有限公司 | A kind of method, apparatus segmented to text sequence and electronic equipment |
CN108108699A (en) * | 2017-12-25 | 2018-06-01 | 重庆邮电大学 | Merge deep neural network model and the human motion recognition method of binary system Hash |
CN108229407A (en) * | 2018-01-11 | 2018-06-29 | 武汉米人科技有限公司 | A kind of behavioral value method and system in video analysis |
CN108304911A (en) * | 2018-01-09 | 2018-07-20 | 中国科学院自动化研究所 | Knowledge Extraction Method and system based on Memory Neural Networks and equipment |
CN108509880A (en) * | 2018-03-21 | 2018-09-07 | 南京邮电大学 | A kind of video personage behavior method for recognizing semantics |
CN108520753A (en) * | 2018-02-26 | 2018-09-11 | 南京工程学院 | Voice lie detection method based on the two-way length of convolution memory network in short-term |
CN108573246A (en) * | 2018-05-08 | 2018-09-25 | 北京工业大学 | A kind of sequential action identification method based on deep learning |
CN108764009A (en) * | 2018-03-21 | 2018-11-06 | 苏州大学 | The Video Events recognition methods of memory network in short-term is grown based on depth residual error |
CN108776779A (en) * | 2018-05-25 | 2018-11-09 | 西安电子科技大学 | SAR Target Recognition of Sequential Images methods based on convolution loop network |
CN108829722A (en) * | 2018-05-08 | 2018-11-16 | 国家计算机网络与信息安全管理中心 | A kind of Dual-Attention relationship classification method and system of remote supervisory |
CN108830305A (en) * | 2018-05-30 | 2018-11-16 | 西南交通大学 | A kind of real-time fire monitoring method of combination DCLRN network and optical flow method |
CN108921047A (en) * | 2018-06-12 | 2018-11-30 | 江西理工大学 | A kind of multi-model ballot mean value action identification method based on cross-layer fusion |
CN108985223A (en) * | 2018-07-12 | 2018-12-11 | 天津艾思科尔科技有限公司 | A kind of human motion recognition method |
CN109255284A (en) * | 2018-07-10 | 2019-01-22 | 西安理工大学 | A kind of Activity recognition method of the 3D convolutional neural networks based on motion profile |
CN109271901A (en) * | 2018-08-31 | 2019-01-25 | 武汉大学 | A kind of sign Language Recognition Method based on Multi-source Information Fusion |
CN109282837A (en) * | 2018-10-24 | 2019-01-29 | 福州大学 | Bragg grating based on LSTM network interlocks the demodulation method of spectrum |
CN109284682A (en) * | 2018-08-21 | 2019-01-29 | 南京邮电大学 | A kind of gesture identification method and system based on STT-LSTM network |
CN109325440A (en) * | 2018-09-19 | 2019-02-12 | 深圳市赢世体育科技有限公司 | Human motion recognition method and system |
CN109344960A (en) * | 2018-09-01 | 2019-02-15 | 哈尔滨工程大学 | A kind of DGRU neural network and its prediction model method for building up preventing data information loss |
CN109426782A (en) * | 2017-08-29 | 2019-03-05 | 北京三星通信技术研究有限公司 | Method for checking object and nerve network system for object detection |
CN109472298A (en) * | 2018-10-19 | 2019-03-15 | 天津大学 | Depth binary feature pyramid for the detection of small scaled target enhances network |
CN109508686A (en) * | 2018-11-26 | 2019-03-22 | 南京邮电大学 | A kind of Human bodys' response method based on the study of stratification proper subspace |
CN109558805A (en) * | 2018-11-06 | 2019-04-02 | 南京邮电大学 | Human bodys' response method based on multilayer depth characteristic |
CN109711380A (en) * | 2019-01-03 | 2019-05-03 | 电子科技大学 | A kind of timing behavior segment generation system and method based on global context information |
CN109753984A (en) * | 2017-11-07 | 2019-05-14 | 北京京东尚科信息技术有限公司 | Video classification methods, device and computer readable storage medium |
CN109753985A (en) * | 2017-11-07 | 2019-05-14 | 北京京东尚科信息技术有限公司 | Video classification methods and device |
CN109753897A (en) * | 2018-12-21 | 2019-05-14 | 西北工业大学 | Based on memory unit reinforcing-time-series dynamics study Activity recognition method |
CN109815785A (en) * | 2018-12-05 | 2019-05-28 | 四川大学 | A kind of face Emotion identification method based on double-current convolutional neural networks |
CN109977904A (en) * | 2019-04-04 | 2019-07-05 | 成都信息工程大学 | A kind of human motion recognition method of the light-type based on deep learning |
CN110110651A (en) * | 2019-04-29 | 2019-08-09 | 齐鲁工业大学 | Activity recognition method in video based on space-time importance and 3D CNN |
CN110245581A (en) * | 2019-05-25 | 2019-09-17 | 天津大学 | A kind of Human bodys' response method based on deep learning and distance-Doppler sequence |
CN110276265A (en) * | 2019-05-27 | 2019-09-24 | 魏运 | Pedestrian monitoring method and device based on intelligent three-dimensional solid monitoring device |
CN110287820A (en) * | 2019-06-06 | 2019-09-27 | 北京清微智能科技有限公司 | Activity recognition method, apparatus, equipment and medium based on LRCN network |
CN110287879A (en) * | 2019-06-26 | 2019-09-27 | 天津大学 | A kind of video behavior recognition methods based on attention mechanism |
CN110287816A (en) * | 2019-06-05 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Car door motion detection method, device and computer readable storage medium |
CN110390294A (en) * | 2019-07-19 | 2019-10-29 | 中国人民解放军国防科技大学 | Target tracking method based on bidirectional long-short term memory neural network |
CN110443182A (en) * | 2019-07-30 | 2019-11-12 | 深圳市博铭维智能科技有限公司 | A kind of urban discharging pipeline video abnormality detection method based on more case-based learnings |
CN110533053A (en) * | 2018-05-23 | 2019-12-03 | 杭州海康威视数字技术股份有限公司 | A kind of event detecting method, device and electronic equipment |
CN110580336A (en) * | 2018-06-08 | 2019-12-17 | 北京得意音通技术有限责任公司 | Lip language word segmentation method and device, storage medium and electronic equipment |
CN110664412A (en) * | 2019-09-19 | 2020-01-10 | 天津师范大学 | Human activity recognition method facing wearable sensor |
CN110751181A (en) * | 2019-09-23 | 2020-02-04 | 华中科技大学 | Target identification method based on sum pooling characteristics |
CN110765845A (en) * | 2019-09-04 | 2020-02-07 | 江苏大学 | Behavior identification method based on video |
CN111027448A (en) * | 2019-12-04 | 2020-04-17 | 成都考拉悠然科技有限公司 | Video behavior category identification method based on time domain inference graph |
CN111052151A (en) * | 2017-10-06 | 2020-04-21 | 高通股份有限公司 | Video motion localization based on attention suggestions |
CN111047901A (en) * | 2019-11-05 | 2020-04-21 | 珠海格力电器股份有限公司 | Parking management method, parking management device, storage medium and computer equipment |
CN111144289A (en) * | 2019-12-26 | 2020-05-12 | 南京航空航天大学 | Identification method for complex human behaviors in video |
CN111241963A (en) * | 2020-01-06 | 2020-06-05 | 中山大学 | First-person visual angle video interactive behavior identification method based on interactive modeling |
CN111259919A (en) * | 2018-11-30 | 2020-06-09 | 杭州海康威视数字技术股份有限公司 | Video classification method, device and equipment and storage medium |
CN111368666A (en) * | 2020-02-25 | 2020-07-03 | 上海蠡图信息科技有限公司 | Living body detection method based on novel pooling and attention mechanism double-current network |
CN111401166A (en) * | 2020-03-06 | 2020-07-10 | 中国科学技术大学 | Robust gesture recognition method based on electromyographic information decoding |
CN111401149A (en) * | 2020-02-27 | 2020-07-10 | 西北工业大学 | Lightweight video behavior identification method based on long-short-term time domain modeling algorithm |
CN111523361A (en) * | 2019-12-26 | 2020-08-11 | 中国科学技术大学 | Human behavior recognition method |
CN112001420A (en) * | 2020-07-24 | 2020-11-27 | 武汉安视感知科技有限公司 | Intelligent timing and counting method and device for drill pipe of mine worker and storage device |
CN112749672A (en) * | 2021-01-19 | 2021-05-04 | 携程旅游网络技术(上海)有限公司 | Photo album video identification method, system, equipment and storage medium |
CN112766420A (en) * | 2021-03-12 | 2021-05-07 | 合肥共达职业技术学院 | Human behavior identification method based on time-frequency domain information |
CN113052885A (en) * | 2021-03-29 | 2021-06-29 | 中国海洋大学 | Underwater environment safety assessment method based on optical flow and depth estimation |
CN113743357A (en) * | 2021-09-16 | 2021-12-03 | 京东科技信息技术有限公司 | Video representation self-supervision contrast learning method and device |
CN114403878A (en) * | 2022-01-20 | 2022-04-29 | 南通理工学院 | Voice fatigue detection method based on deep learning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104217214A (en) * | 2014-08-21 | 2014-12-17 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | Configurable convolutional neural network based red green blue-distance (RGB-D) figure behavior identification method |
CN104615983A (en) * | 2015-01-28 | 2015-05-13 | 中国科学院自动化研究所 | Behavior identification method based on recurrent neural network and human skeleton movement sequences |
CN105095862A (en) * | 2015-07-10 | 2015-11-25 | 南开大学 | Human gesture recognizing method based on depth convolution condition random field |
CN105354572A (en) * | 2015-12-10 | 2016-02-24 | 苏州大学 | Automatic identification system of number plate on the basis of simplified convolutional neural network |
-
2016
- 2016-12-21 CN CN201611193290.1A patent/CN106845351A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104217214A (en) * | 2014-08-21 | 2014-12-17 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | Configurable convolutional neural network based red green blue-distance (RGB-D) figure behavior identification method |
CN104615983A (en) * | 2015-01-28 | 2015-05-13 | 中国科学院自动化研究所 | Behavior identification method based on recurrent neural network and human skeleton movement sequences |
CN105095862A (en) * | 2015-07-10 | 2015-11-25 | 南开大学 | Human gesture recognizing method based on depth convolution condition random field |
CN105354572A (en) * | 2015-12-10 | 2016-02-24 | 苏州大学 | Automatic identification system of number plate on the basis of simplified convolutional neural network |
Non-Patent Citations (2)
Title |
---|
JEFF DONAHUE ET AL;: "《Long-term Recurrent Convolutional Networks for Visual Recognition and Description》", 《2015 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
KAREN SIMONYAN: "《Two-Stream Convolutional Networks》", 《ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS》 * |
Cited By (96)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107330463A (en) * | 2017-06-29 | 2017-11-07 | 南京信息工程大学 | Model recognizing method based on CNN multiple features combinings and many nuclear sparse expressions |
CN107463879A (en) * | 2017-07-05 | 2017-12-12 | 成都数联铭品科技有限公司 | Human bodys' response method based on deep learning |
CN107463949B (en) * | 2017-07-14 | 2020-02-21 | 北京协同创新研究院 | Video action classification processing method and device |
CN107463949A (en) * | 2017-07-14 | 2017-12-12 | 北京协同创新研究院 | A kind of processing method and processing device of video actions classification |
CN107484017A (en) * | 2017-07-25 | 2017-12-15 | 天津大学 | Supervision video abstraction generating method is had based on attention model |
CN107484017B (en) * | 2017-07-25 | 2020-05-26 | 天津大学 | Supervised video abstract generation method based on attention model |
CN107423721A (en) * | 2017-08-08 | 2017-12-01 | 珠海习悦信息技术有限公司 | Interactive action detection method, device, storage medium and processor |
CN107506712A (en) * | 2017-08-15 | 2017-12-22 | 成都考拉悠然科技有限公司 | Method for distinguishing is known in a kind of human behavior based on 3D depth convolutional networks |
CN107506712B (en) * | 2017-08-15 | 2021-05-18 | 成都考拉悠然科技有限公司 | Human behavior identification method based on 3D deep convolutional network |
CN109426782A (en) * | 2017-08-29 | 2019-03-05 | 北京三星通信技术研究有限公司 | Method for checking object and nerve network system for object detection |
CN109426782B (en) * | 2017-08-29 | 2023-09-19 | 北京三星通信技术研究有限公司 | Object detection method and neural network system for object detection |
CN111052151B (en) * | 2017-10-06 | 2023-07-11 | 高通股份有限公司 | Video action positioning based on attention suggestion |
CN111052151A (en) * | 2017-10-06 | 2020-04-21 | 高通股份有限公司 | Video motion localization based on attention suggestions |
CN107644519A (en) * | 2017-10-09 | 2018-01-30 | 中电科新型智慧城市研究院有限公司 | A kind of intelligent alarm method and system based on video human Activity recognition |
CN107679522B (en) * | 2017-10-31 | 2020-10-13 | 内江师范学院 | Multi-stream LSTM-based action identification method |
CN107679522A (en) * | 2017-10-31 | 2018-02-09 | 内江师范学院 | Action identification method based on multithread LSTM |
CN109753984A (en) * | 2017-11-07 | 2019-05-14 | 北京京东尚科信息技术有限公司 | Video classification methods, device and computer readable storage medium |
CN109753985A (en) * | 2017-11-07 | 2019-05-14 | 北京京东尚科信息技术有限公司 | Video classification methods and device |
CN108009493B (en) * | 2017-11-30 | 2021-07-06 | 电子科技大学 | Human face anti-cheating recognition method based on motion enhancement |
CN108009493A (en) * | 2017-11-30 | 2018-05-08 | 电子科技大学 | Face anti-fraud recognition methods based on action enhancing |
CN108038103B (en) * | 2017-12-18 | 2021-08-10 | 沈阳智能大数据科技有限公司 | Method and device for segmenting text sequence and electronic equipment |
CN108038103A (en) * | 2017-12-18 | 2018-05-15 | 北京百分点信息科技有限公司 | A kind of method, apparatus segmented to text sequence and electronic equipment |
CN108108699A (en) * | 2017-12-25 | 2018-06-01 | 重庆邮电大学 | Merge deep neural network model and the human motion recognition method of binary system Hash |
CN108304911A (en) * | 2018-01-09 | 2018-07-20 | 中国科学院自动化研究所 | Knowledge Extraction Method and system based on Memory Neural Networks and equipment |
CN108229407A (en) * | 2018-01-11 | 2018-06-29 | 武汉米人科技有限公司 | A kind of behavioral value method and system in video analysis |
CN108520753A (en) * | 2018-02-26 | 2018-09-11 | 南京工程学院 | Voice lie detection method based on the two-way length of convolution memory network in short-term |
CN108520753B (en) * | 2018-02-26 | 2020-07-24 | 南京工程学院 | Voice lie detection method based on convolution bidirectional long-time and short-time memory network |
CN108764009A (en) * | 2018-03-21 | 2018-11-06 | 苏州大学 | The Video Events recognition methods of memory network in short-term is grown based on depth residual error |
CN108509880A (en) * | 2018-03-21 | 2018-09-07 | 南京邮电大学 | A kind of video personage behavior method for recognizing semantics |
CN108829722A (en) * | 2018-05-08 | 2018-11-16 | 国家计算机网络与信息安全管理中心 | A kind of Dual-Attention relationship classification method and system of remote supervisory |
CN108573246B (en) * | 2018-05-08 | 2022-04-05 | 北京工业大学 | Time sequence action identification method based on deep learning |
CN108829722B (en) * | 2018-05-08 | 2020-10-02 | 国家计算机网络与信息安全管理中心 | Remote supervision Dual-Attention relation classification method and system |
CN108573246A (en) * | 2018-05-08 | 2018-09-25 | 北京工业大学 | A kind of sequential action identification method based on deep learning |
CN110533053A (en) * | 2018-05-23 | 2019-12-03 | 杭州海康威视数字技术股份有限公司 | A kind of event detecting method, device and electronic equipment |
CN108776779A (en) * | 2018-05-25 | 2018-11-09 | 西安电子科技大学 | SAR Target Recognition of Sequential Images methods based on convolution loop network |
CN108830305A (en) * | 2018-05-30 | 2018-11-16 | 西南交通大学 | A kind of real-time fire monitoring method of combination DCLRN network and optical flow method |
CN110580336A (en) * | 2018-06-08 | 2019-12-17 | 北京得意音通技术有限责任公司 | Lip language word segmentation method and device, storage medium and electronic equipment |
CN108921047A (en) * | 2018-06-12 | 2018-11-30 | 江西理工大学 | A kind of multi-model ballot mean value action identification method based on cross-layer fusion |
CN108921047B (en) * | 2018-06-12 | 2021-11-26 | 江西理工大学 | Multi-model voting mean value action identification method based on cross-layer fusion |
CN109255284A (en) * | 2018-07-10 | 2019-01-22 | 西安理工大学 | A kind of Activity recognition method of the 3D convolutional neural networks based on motion profile |
CN108985223A (en) * | 2018-07-12 | 2018-12-11 | 天津艾思科尔科技有限公司 | A kind of human motion recognition method |
CN109284682A (en) * | 2018-08-21 | 2019-01-29 | 南京邮电大学 | A kind of gesture identification method and system based on STT-LSTM network |
CN109271901A (en) * | 2018-08-31 | 2019-01-25 | 武汉大学 | A kind of sign Language Recognition Method based on Multi-source Information Fusion |
CN109344960A (en) * | 2018-09-01 | 2019-02-15 | 哈尔滨工程大学 | A kind of DGRU neural network and its prediction model method for building up preventing data information loss |
CN109325440A (en) * | 2018-09-19 | 2019-02-12 | 深圳市赢世体育科技有限公司 | Human motion recognition method and system |
CN109325440B (en) * | 2018-09-19 | 2021-04-30 | 深圳市赢世体育科技有限公司 | Human body action recognition method and system |
CN109472298B (en) * | 2018-10-19 | 2021-06-01 | 天津大学 | Deep bidirectional feature pyramid enhanced network for small-scale target detection |
CN109472298A (en) * | 2018-10-19 | 2019-03-15 | 天津大学 | Depth binary feature pyramid for the detection of small scaled target enhances network |
CN109282837A (en) * | 2018-10-24 | 2019-01-29 | 福州大学 | Bragg grating based on LSTM network interlocks the demodulation method of spectrum |
CN109558805A (en) * | 2018-11-06 | 2019-04-02 | 南京邮电大学 | Human bodys' response method based on multilayer depth characteristic |
CN109508686A (en) * | 2018-11-26 | 2019-03-22 | 南京邮电大学 | A kind of Human bodys' response method based on the study of stratification proper subspace |
CN111259919A (en) * | 2018-11-30 | 2020-06-09 | 杭州海康威视数字技术股份有限公司 | Video classification method, device and equipment and storage medium |
CN111259919B (en) * | 2018-11-30 | 2024-01-23 | 杭州海康威视数字技术股份有限公司 | Video classification method, device and equipment and storage medium |
CN109815785A (en) * | 2018-12-05 | 2019-05-28 | 四川大学 | A kind of face Emotion identification method based on double-current convolutional neural networks |
CN109753897A (en) * | 2018-12-21 | 2019-05-14 | 西北工业大学 | Based on memory unit reinforcing-time-series dynamics study Activity recognition method |
CN109753897B (en) * | 2018-12-21 | 2022-05-27 | 西北工业大学 | Behavior recognition method based on memory cell reinforcement-time sequence dynamic learning |
CN109711380A (en) * | 2019-01-03 | 2019-05-03 | 电子科技大学 | A kind of timing behavior segment generation system and method based on global context information |
CN109977904A (en) * | 2019-04-04 | 2019-07-05 | 成都信息工程大学 | A kind of human motion recognition method of the light-type based on deep learning |
CN110110651A (en) * | 2019-04-29 | 2019-08-09 | 齐鲁工业大学 | Activity recognition method in video based on space-time importance and 3D CNN |
CN110110651B (en) * | 2019-04-29 | 2023-06-13 | 齐鲁工业大学 | Method for identifying behaviors in video based on space-time importance and 3D CNN |
CN110245581B (en) * | 2019-05-25 | 2023-04-07 | 天津大学 | Human behavior recognition method based on deep learning and distance-Doppler sequence |
CN110245581A (en) * | 2019-05-25 | 2019-09-17 | 天津大学 | A kind of Human bodys' response method based on deep learning and distance-Doppler sequence |
CN110276265A (en) * | 2019-05-27 | 2019-09-24 | 魏运 | Pedestrian monitoring method and device based on intelligent three-dimensional solid monitoring device |
CN110287816A (en) * | 2019-06-05 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Car door motion detection method, device and computer readable storage medium |
CN110287820A (en) * | 2019-06-06 | 2019-09-27 | 北京清微智能科技有限公司 | Activity recognition method, apparatus, equipment and medium based on LRCN network |
CN110287820B (en) * | 2019-06-06 | 2021-07-23 | 北京清微智能科技有限公司 | Behavior recognition method, device, equipment and medium based on LRCN network |
CN110287879A (en) * | 2019-06-26 | 2019-09-27 | 天津大学 | A kind of video behavior recognition methods based on attention mechanism |
CN110390294B (en) * | 2019-07-19 | 2021-03-09 | 中国人民解放军国防科技大学 | Target tracking method based on bidirectional long-short term memory neural network |
CN110390294A (en) * | 2019-07-19 | 2019-10-29 | 中国人民解放军国防科技大学 | Target tracking method based on bidirectional long-short term memory neural network |
CN110443182A (en) * | 2019-07-30 | 2019-11-12 | 深圳市博铭维智能科技有限公司 | A kind of urban discharging pipeline video abnormality detection method based on more case-based learnings |
CN110765845B (en) * | 2019-09-04 | 2023-08-22 | 江苏大学 | Behavior recognition method based on video |
CN110765845A (en) * | 2019-09-04 | 2020-02-07 | 江苏大学 | Behavior identification method based on video |
CN110664412A (en) * | 2019-09-19 | 2020-01-10 | 天津师范大学 | Human activity recognition method facing wearable sensor |
CN110751181A (en) * | 2019-09-23 | 2020-02-04 | 华中科技大学 | Target identification method based on sum pooling characteristics |
CN111047901A (en) * | 2019-11-05 | 2020-04-21 | 珠海格力电器股份有限公司 | Parking management method, parking management device, storage medium and computer equipment |
CN111027448A (en) * | 2019-12-04 | 2020-04-17 | 成都考拉悠然科技有限公司 | Video behavior category identification method based on time domain inference graph |
CN111523361B (en) * | 2019-12-26 | 2022-09-06 | 中国科学技术大学 | Human behavior recognition method |
CN111523361A (en) * | 2019-12-26 | 2020-08-11 | 中国科学技术大学 | Human behavior recognition method |
CN111144289B (en) * | 2019-12-26 | 2024-03-22 | 南京航空航天大学 | Identification method for complex human behaviors in video |
CN111144289A (en) * | 2019-12-26 | 2020-05-12 | 南京航空航天大学 | Identification method for complex human behaviors in video |
CN111241963A (en) * | 2020-01-06 | 2020-06-05 | 中山大学 | First-person visual angle video interactive behavior identification method based on interactive modeling |
CN111241963B (en) * | 2020-01-06 | 2023-07-14 | 中山大学 | First person view video interactive behavior identification method based on interactive modeling |
CN111368666A (en) * | 2020-02-25 | 2020-07-03 | 上海蠡图信息科技有限公司 | Living body detection method based on novel pooling and attention mechanism double-current network |
CN111368666B (en) * | 2020-02-25 | 2023-08-18 | 上海蠡图信息科技有限公司 | Living body detection method based on novel pooling and attention mechanism double-flow network |
CN111401149B (en) * | 2020-02-27 | 2022-05-13 | 西北工业大学 | Lightweight video behavior identification method based on long-short-term time domain modeling algorithm |
CN111401149A (en) * | 2020-02-27 | 2020-07-10 | 西北工业大学 | Lightweight video behavior identification method based on long-short-term time domain modeling algorithm |
CN111401166A (en) * | 2020-03-06 | 2020-07-10 | 中国科学技术大学 | Robust gesture recognition method based on electromyographic information decoding |
CN112001420B (en) * | 2020-07-24 | 2022-09-09 | 武汉安视感知科技有限公司 | Intelligent timing and counting method and device for drill pipe of mine worker and storage device |
CN112001420A (en) * | 2020-07-24 | 2020-11-27 | 武汉安视感知科技有限公司 | Intelligent timing and counting method and device for drill pipe of mine worker and storage device |
CN112749672A (en) * | 2021-01-19 | 2021-05-04 | 携程旅游网络技术(上海)有限公司 | Photo album video identification method, system, equipment and storage medium |
CN112766420B (en) * | 2021-03-12 | 2022-10-21 | 合肥共达职业技术学院 | Human behavior identification method based on time-frequency domain information |
CN112766420A (en) * | 2021-03-12 | 2021-05-07 | 合肥共达职业技术学院 | Human behavior identification method based on time-frequency domain information |
CN113052885A (en) * | 2021-03-29 | 2021-06-29 | 中国海洋大学 | Underwater environment safety assessment method based on optical flow and depth estimation |
CN113743357B (en) * | 2021-09-16 | 2023-12-05 | 京东科技信息技术有限公司 | Video characterization self-supervision contrast learning method and device |
CN113743357A (en) * | 2021-09-16 | 2021-12-03 | 京东科技信息技术有限公司 | Video representation self-supervision contrast learning method and device |
CN114403878A (en) * | 2022-01-20 | 2022-04-29 | 南通理工学院 | Voice fatigue detection method based on deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106845351A (en) | It is a kind of for Activity recognition method of the video based on two-way length mnemon in short-term | |
CN104318558B (en) | Hand Gesture Segmentation method based on Multi-information acquisition under complex scene | |
CN106570477B (en) | Vehicle cab recognition model building method and model recognizing method based on deep learning | |
CN110427839A (en) | Video object detection method based on multilayer feature fusion | |
CN106407889A (en) | Video human body interaction motion identification method based on optical flow graph depth learning model | |
CN110378259A (en) | A kind of multiple target Activity recognition method and system towards monitor video | |
CN107609460A (en) | A kind of Human bodys' response method for merging space-time dual-network stream and attention mechanism | |
CN107016357A (en) | A kind of video pedestrian detection method based on time-domain convolutional neural networks | |
CN103226708B (en) | A kind of multi-model fusion video hand division method based on Kinect | |
CN107767405A (en) | A kind of nuclear phase for merging convolutional neural networks closes filtered target tracking | |
CN106951867A (en) | Face identification method, device, system and equipment based on convolutional neural networks | |
CN104281853A (en) | Behavior identification method based on 3D convolution neural network | |
CN109376747A (en) | A kind of video flame detecting method based on double-current convolutional neural networks | |
CN106909938B (en) | Visual angle independence behavior identification method based on deep learning network | |
CN106485214A (en) | A kind of eyes based on convolutional neural networks and mouth state identification method | |
CN106156765A (en) | safety detection method based on computer vision | |
Yu et al. | Railway obstacle detection algorithm using neural network | |
CN110532925B (en) | Driver fatigue detection method based on space-time graph convolutional network | |
CN107301376B (en) | Pedestrian detection method based on deep learning multi-layer stimulation | |
CN108648211A (en) | A kind of small target detecting method, device, equipment and medium based on deep learning | |
CN106127812A (en) | A kind of passenger flow statistical method of non-gate area, passenger station based on video monitoring | |
CN110909672A (en) | Smoking action recognition method based on double-current convolutional neural network and SVM | |
Xu et al. | An efficient foreign objects detection network for power substation | |
CN107992854A (en) | Forest Ecology man-machine interaction method based on machine vision | |
Yan | RETRACTED ARTICLE: Researches on hybrid algorithm for moving target detection and tracking in sports video |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170613 |