CN108830262A - Multi-angle human face expression recognition method under natural conditions - Google Patents
Multi-angle human face expression recognition method under natural conditions Download PDFInfo
- Publication number
- CN108830262A CN108830262A CN201810830347.7A CN201810830347A CN108830262A CN 108830262 A CN108830262 A CN 108830262A CN 201810830347 A CN201810830347 A CN 201810830347A CN 108830262 A CN108830262 A CN 108830262A
- Authority
- CN
- China
- Prior art keywords
- layer
- image
- angle
- human face
- expression recognition
- 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
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2431—Multiple classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/174—Facial expression recognition
- G06V40/175—Static expression
Abstract
The present invention relates to a kind of Multi-angle human face expression recognition methods under natural conditions, input the Facial Expression Image data of multi-angle, pre-process to its image data;MVFE-LightNet network structure is established, after pretreatment after image data input input layer, using two two-dimensional convolution layers, extracts the rudimentary edge feature of image;Again successively after the separable convolutional layer of 6 residual error depth, 1 two-dimensional convolution layer and 1 GlobalAveragePooling2D layers of progress image further feature extract and process;The image further feature finally extracted is sent into Softmax classifier and is trained and identifies, final classification output.Method is suitable for the facial expression recognition under multi-angle;This method arithmetic speed is fast, can be realized real-time operation;This method all has higher discrimination in different angle facial expression recognition;This method efficiently solves the network number of plies and increases and bring overfitting problem.
Description
Technical field
The present invention relates to a kind of face recognition technology, in particular to a kind of natural shape based on MVFE-LightNet network
Multi-angle human face expression recognition method under state.
Background technique
Facial expression recognition is one of the research hotspot of computer vision and area of pattern recognition, is human-computer interaction and emotion
The development trend of computing technique research obtains widely in intelligent security guard, robot building, medical treatment, communication and automotive field etc.
Using.The emotion of people is largely presented in human face expression, and the underlying thought of people can be judged by the variation of expression.
With the development of artificial intelligence, demand of the mankind to the life of intelligent and comfortableization increasingly increases, and obtains people by human face expression
The analysis of class emotion information not only has scientific research value, is also of great significance to human psychology state and affective comprehension.
In the past few decades, the research achievement of facial expression recognition is mainly for front or nearly front face image.
And multi-orientation Face expression under natural conditions obviously has wider application field and higher application value.With positive dough figurine
For face expression compared to identification, non-frontal face needs to handle human face posture variation bring expression information missing, multi-pose feature
With the problems such as, substantially increase face acquisition, detection and identification difficulty.
Conventional face's Expression Recognition algorithm is generally divided into two steps:Feature extraction and classifier differentiate.Wherein, feature mentions
Method is taken to have special based on appearance or geometrical characteristic, histograms of oriented gradients (HOG), discrete cosine transform (DCT) and Scale invariant
(main includes spy for sign transformation (SIFT) etc. and their different variants, Feature Dimension Reduction (PCA, LDA and SVD) and Fusion Features
Levy grade fusion and Decision-level fusion).For classifier, the classifier of most of prevalences, such as support vector machines (SVM) and shellfish
This classifier of leaf, together with some unsupervised learning arts.But this kind of algorithm needs handmarking's characteristic point, it cannot be according to class
With Image Adjusting feature extraction, lack robustness and practicability.If selected extraction characterization method, which lacks, distinguishes classification institute
Need characterization ability, then the accuracy of disaggregated model will receive very big influence, to a certain extent with used classification policy
Type it is unrelated.
Summary of the invention
The present invention be directed to the problems of non-frontal human face detection and recognition hardly possible, propose a kind of Multi-angle human under natural conditions
Face expression recognition method, it can be achieved that human face posture variation greatly, the serious facial expression recognition of facial information missing.
The technical scheme is that:A kind of Multi-angle human face expression recognition method under natural conditions, specifically includes as follows
Step:
1) the Facial Expression Image data for inputting multi-angle, pre-process its image data, include the following steps:
1.1) coarse localization of detection and key point is carried out to face using multitask concatenated convolutional neural network method,
Human face region is cut again;
1.2) global contrast normalizes:To have already passed through Face datection and cut after original image carry out gray processing and
Normalized controls the scale of each feature in corresponding range;
1.3) human face expression data expand:" enhancing " is carried out to image using stochastic transformation to handle, and picture is carried out automatically
Amplification data;
2) MVFE-LightNet network frame is established:MVFE-LightNet network structure successively includes input layer, two
Two-dimensional convolution layer, the separable convolutional layer of 6 residual error depth, 1 two-dimensional convolution layer, 1 GlobalAveragePooling2D
Layer and Softmax classifier output layer;After pretreatment after image data input input layer, using two two-dimensional convolution layers, mention
Take the rudimentary edge feature of image;Image successively is realized by the separable convolutional layer of 6 residual error depth, 1 two-dimensional convolution layer again
Further feature is extracted, and image further feature passes through 1 GlobalAveragePooling2D layers again and applies the overall situation to airspace signal
Average value pond after the replacement connected entirely, is sent into Softmax classifier and is trained and identifies, final classification output.
One Batch Normalization function of each convolution heel and a ReLU activate letter in the step 2)
Number, wherein along with a MaxPooling2D calculating after the separable convolutional layer of residual error depth.Beneficial effects of the present invention exist
In:Present invention Multi-angle human face expression recognition method under natural conditions, suitable for the facial expression recognition under multi-angle;This method
Arithmetic speed is fast, can be realized real-time operation;This method all has higher discrimination in different angle facial expression recognition;
This method efficiently solves the network number of plies and increases and bring overfitting problem.
Detailed description of the invention
Fig. 1 is multi-view face detection of the present invention and cutting schematic diagram;
Fig. 2 is multi-angle of view human face expression gray processing of the present invention and normalization schematic diagram;
Fig. 3 is that face expression data of the present invention expands schematic diagram;
Fig. 4 is that the present invention is based on MVFE-LightNet network model schematic diagrames;
Fig. 5 is that Conv2D convolution sample of the present invention exports schematic diagram;
Fig. 6 is that image deep layer convolution of the present invention extracts image further feature schematic diagram;
Fig. 7 is the experimental result picture of the multi-angle of view facial expression recognition confusion matrix in BU-3DFE database of the present invention.
Specific embodiment
1, human face expression data prediction
The Facial Expression Image data for inputting multi-angle, pre-process its image data, if Fig. 1,2,3 are establishing
The schematic diagram of human face expression data prediction.
(1) Face datection and cutting:Using multitask concatenated convolutional neural network (Multi-Task Convolutional
Neural Net-work, MTCNN) method carries out the coarse localization of detection and key point to face, then human face region is cut,
Multi-view face detection as shown in Figure 1 and cutting schematic diagram.MTCNN is the nerve net of a deep layer grade (24 layers) multitask frame
Network, firstly, adjustment image size constructs image pyramid, and the input as three-stage cascade frame to different scales, three
The cascade structure of the deep convolutional network of grade is as follows:
P-Net(Proposal Network):Candidate face frame and its bounding box regression vector are generated, the side is then used
Boundary's frame regression vector calibrates candidate frame, and merges the candidate frame of high superposed using non-maxima suppression (NMS) method.
R-Net(Refine Network):Candidate frame is further screened, is corrected using bounding box recurrence, and use NMS
Merge candidate frame.
O-Net(Output Network):It is similar with R-Net, best candidate frame is selected, and export the position of five characteristic points
It sets.
(2) global contrast normalizes:Normalization is that the original image after having already passed through Face datection and cutting is each
The scale of feature is controlled in corresponding range.Facial image carries out gray processing and normalized, multi-angle of view face as shown in Figure 2
Expression gray processing and normalization schematic diagram, with frontal one be 90 degree indicates, other angles expression 0 °, 45 °, 90 °, 135 ° with
180°.In addition to neutrality, including anger, detest, frightened, glad, sad and surprised seven kinds of basic facial expressions.
(3) data enhance
It is performed better than when deep learning model treatment large data collection.Image " increase using a series of stochastic transformations
It handles by force ", so that model not can be appreciated that identical image twice, effectively improves picture utilization rate.For example, rotate, overturn,
The transformation such as zooming and panning.Present invention uses width and height displacement, image overall width or height are 0.2, Random-Rotation model
Enclosing is 0-20 °, and shearing range is 0.1, and zooming range is [0-0.1].Also image level is overturn, and applies " fill pattern " plan
Newly created pixel is slightly filled, using the picture pretreating tool of deep learning frame keras automatically to the amplification data of picture,
Human face expression data as shown in Figure 3 expand schematic diagram.
2, MVFE-LightNet network frame
The basic structure of the MVFE-LightNet network model that the present invention designs as shown in figure 4, with Xception and
Based on Resnet framework, this separable convolution sum residual error network module of architecture combined depth, network parameter is reduced same
When, network performance is not lost, is trained using Adam optimizer.
MVFE-LightNet network structure is a full convolutional neural networks, and MVFE-LightNet network structure is successively wrapped
Include input layer, two two-dimensional convolution layers, the separable convolutional layer of 6 residual error depth, 1 two-dimensional convolution layer, 1
GlobalAveragePooling2D layers and 1 Softmax output layer.In addition to GlobalAveragePooling2D layers last,
One Batch Normalization function of each convolutional layer heel and a ReLU activation primitive, wherein each residual error depth
A MaxPooling2D operation is added after separable convolutional layer.Input layer is that input size is 64 × 64 × 1 pixel
Multi-angle facial expression image, two two-dimensional convolution layers (Conv2D) use size to carry out sliding window volume for 3 × 38 convolution kernels
Product, this layer extract the rudimentary edge feature of image, remain the details of image, Conv2D convolution sample output signal as shown in Figure 5
Figure.The image further feature of extraction image deep layer convolution as shown in Figure 6 extracts image further feature schematic diagram, a-f difference in Fig. 6
The output that the 1st to the 6th residual error depth separates convolutional layer, it can be found that more toward deep layer, expressed by information it is more abstract multiple
It is miscellaneous.The last layer is that airspace signal applies global mean value pond with GlobalAveragePooling2D layers, is to be connected entirely
The replacement connect reduces the quantity of parameter, and doing regularization in structure to whole network prevents over-fitting.Output layer is Softmax
Classifier predicts that GlobalAveragePooling2D layers of output are sent into Softmax classifiers and are trained and identify to generate,
It is finally reached classification purpose.This is popularization of the Logic Regression Models in more classification problems, and in more classification problems, k can be predicted
Kind may (species number of the k for sample label, k=6 herein, angle be divided into 5 angles, are the faces for being directed to 5 angles respectively
What expression was trained and identified, that is, classifier do not distinguish which angle is current expression be.In other words, we are preparatory
Known facial angle is then input in the classifier of some angle and carries out Expression Recognition), it is assumed that classifier input feature vector is(Refer to n+1 dimension space.N+1 is an assumption value, that is, passes through 6 layers of separable convolution of residual error depth
Output feature vector hypothesis after layer is n+1 dimension, then inputs to softmax and classifies), sample label y(i), that is, constitute
Classification layer supervised learning training set S={ (x(1), y(1)), (x(2), y(2)) ..., (x(m), y(m)), it is assumed that function
hθ(x) and cost function J (θ) form difference is as follows:
Wherein P is probability, conditional probability;θ is classifier parameters, has k, each n+1 dimension;It is
Model parameter,For item is normalized to probability distribution, so that the sum of all probability are 1.
Wherein, 1 { }=1 is an indicative function, and value rule is:When expression formula is true in braces, the function
Result be just 1, otherwise its result is just 0.
Softmax regression model is popularization of the logistic regression model in more classification problems, when number of classifying is 2
Time can degenerate for Logistic classification.In more classification problems, class label y can take more than two values.
The hypothesis function of logistic regression model is as follows:
Training pattern parameter θ can minimize cost function J (θ):
X is inputted for given test, with assume function for each classification y=j estimate probability value p (y=j |
X), that is, estimate the probability that each classification results of x occur.Assuming that function will export the vector of k dimension to indicate this k
The probability value of a estimation.Assuming that function hθ(x) form is as follows:
WhereinIt is model parameter,For item is normalized to probability distribution, make
Obtaining the sum of all probability is 1.
For convenience's sake, we equally indicate whole model parameters using symbol theta.Realizing Softmax recurrence
When, θ is indicated with a k × (n+1) matrix can easily, which is by θ1, θ2..., θkIt is enumerated by row
It arrives, as follows:
T is transposition.
In softmax regression algorithm, returning cost function is:
Wherein, 1 { }=1 is an indicative function, and value rule is:When expression formula is true in braces, the function
Result be just 1, otherwise its result is just 0.
3. experimental result and analysis
The discrimination of every kind of expression in BU-3DFE database under different angle is as shown in table 1, between this 6 kinds of expressions
Confusion matrix it is as shown in Figure 7.From table 1 it follows that positive, discrimination 0.837 higher than the discrimination of other angles, this
Outside, whole discrimination is up to 0.887;It can be seen that in six kinds of expressions from these confusion matrixs of Fig. 7, it is surprised and glad
Expression than detest and it is frightened be easier identified, be most likely to be the muscle deformation of both expressions relatively than other expressions
Greatly.
Table 1
Network parameter size as shown in table 2 and training speed compare, and compare the size of several models, discrimination and speed
Degree, it can be seen that the discrimination of MVFE-LightNet network model is apparently higher than LightNet and Mnist-cnn network model,
And there is overfitting problem in Xcepton model.The discrimination of this paper network model is slightly below Resent-18 network model, still
Runing time is about its half, and experimental period is greatly saved.
Table 2
Claims (2)
1. a kind of Multi-angle human face expression recognition method under natural conditions, which is characterized in that specifically comprise the following steps:
1) the Facial Expression Image data for inputting multi-angle, pre-process its image data, include the following steps:
1.1) coarse localization of detection and key point is carried out to face using multitask concatenated convolutional neural network method, then is cut out
Cut human face region;
1.2) global contrast normalizes:Gray processing and normalizing are carried out to the original image after having already passed through Face datection and cutting
Change processing controls the scale of each feature in corresponding range;
1.3) human face expression data expand:" enhancing " is carried out to image using stochastic transformation to handle, and picture is expanded automatically
Data;
2) MVFE-LightNet network frame is established:MVFE-LightNet network structure successively includes input layer, two two dimensions
Convolutional layer, the separable convolutional layer of 6 residual error depth, 1 two-dimensional convolution layer, 1 GlobalAveragePooling2D layers and
Softmax classifier output layer;After pretreatment after image data input input layer, using two two-dimensional convolution layers, figure is extracted
As rudimentary edge feature;Image deep layer successively is realized by the separable convolutional layer of 6 residual error depth, 1 two-dimensional convolution layer again
Feature extraction, image further feature pass through 1 GlobalAveragePooling2D layers again and apply the overall situation averagely to airspace signal
It is worth pond, after the replacement connected entirely, is sent into Softmax classifier and is trained and identifies, final classification output.
2. Multi-angle human face expression recognition method under natural conditions according to claim 1, which is characterized in that the step 2)
In one Batch Normalization function of each convolution heel and a ReLU activation primitive, wherein residual error depth can divide
From convolutional layer after calculate along with MaxPooling2D.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810830347.7A CN108830262A (en) | 2018-07-25 | 2018-07-25 | Multi-angle human face expression recognition method under natural conditions |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810830347.7A CN108830262A (en) | 2018-07-25 | 2018-07-25 | Multi-angle human face expression recognition method under natural conditions |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108830262A true CN108830262A (en) | 2018-11-16 |
Family
ID=64140569
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810830347.7A Pending CN108830262A (en) | 2018-07-25 | 2018-07-25 | Multi-angle human face expression recognition method under natural conditions |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108830262A (en) |
Cited By (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109754017A (en) * | 2019-01-09 | 2019-05-14 | 西北工业大学 | Based on separable three-dimensional residual error network and transfer learning hyperspectral image classification method |
CN109784154A (en) * | 2018-12-10 | 2019-05-21 | 平安科技(深圳)有限公司 | Emotion identification method, apparatus, equipment and medium based on deep neural network |
CN109800648A (en) * | 2018-12-18 | 2019-05-24 | 北京英索科技发展有限公司 | Face datection recognition methods and device based on the correction of face key point |
CN109829409A (en) * | 2019-01-23 | 2019-05-31 | 深兰科技(上海)有限公司 | Driver's emotional state detection method and system |
CN109948509A (en) * | 2019-03-11 | 2019-06-28 | 成都旷视金智科技有限公司 | Obj State monitoring method, device and electronic equipment |
CN110175504A (en) * | 2019-04-08 | 2019-08-27 | 杭州电子科技大学 | A kind of target detection and alignment schemes based on multitask concatenated convolutional network |
CN110188615A (en) * | 2019-04-30 | 2019-08-30 | 中国科学院计算技术研究所 | A kind of facial expression recognizing method, device, medium and system |
CN110210432A (en) * | 2019-06-06 | 2019-09-06 | 湖南大学 | A kind of face identification method based on intelligent security guard robot under the conditions of untethered |
CN110222565A (en) * | 2019-04-26 | 2019-09-10 | 合肥进毅智能技术有限公司 | A kind of method for detecting human face, device, electronic equipment and storage medium |
CN110222559A (en) * | 2019-04-24 | 2019-09-10 | 深圳市微纳集成电路与系统应用研究院 | Smog image detecting method and device based on convolutional neural networks |
CN110580445A (en) * | 2019-07-12 | 2019-12-17 | 西北工业大学 | Face key point detection method based on GIoU and weighted NMS improvement |
CN110660046A (en) * | 2019-08-30 | 2020-01-07 | 太原科技大学 | Industrial product defect image classification method based on lightweight deep neural network |
CN110825852A (en) * | 2019-11-07 | 2020-02-21 | 四川长虹电器股份有限公司 | Long text-oriented semantic matching method and system |
CN110889672A (en) * | 2019-11-19 | 2020-03-17 | 哈尔滨理工大学 | Student card punching and class taking state detection system based on deep learning |
CN111126173A (en) * | 2019-12-04 | 2020-05-08 | 玉林师范学院 | High-precision face detection method |
CN111414884A (en) * | 2020-03-27 | 2020-07-14 | 南京工业大学 | Facial expression recognition method based on edge calculation |
CN111652146A (en) * | 2020-06-03 | 2020-09-11 | 陕西科技大学 | Detection method for facial emotional conditions of old people in nursing home |
CN111695513A (en) * | 2020-06-12 | 2020-09-22 | 长安大学 | Facial expression recognition method based on depth residual error network |
CN111723709A (en) * | 2020-06-09 | 2020-09-29 | 大连海事大学 | Fly face recognition method based on deep convolutional neural network |
CN112215157A (en) * | 2020-10-13 | 2021-01-12 | 北京中电兴发科技有限公司 | Multi-model fusion-based face feature dimension reduction extraction method |
CN112304435A (en) * | 2020-10-10 | 2021-02-02 | 广州中大数字家庭工程技术研究中心有限公司 | Human body thermal imaging temperature measurement method combining face recognition |
CN112487855A (en) * | 2019-09-12 | 2021-03-12 | 上海齐感电子信息科技有限公司 | MTCNN (multiple-connectivity neural network) model-based face detection method and device and terminal |
CN112580458A (en) * | 2020-12-10 | 2021-03-30 | 中国地质大学(武汉) | Facial expression recognition method, device, equipment and storage medium |
CN113065460A (en) * | 2021-03-31 | 2021-07-02 | 吉林农业大学 | Establishment method of pig face facial expression recognition framework based on multitask cascade |
CN113160114A (en) * | 2021-01-29 | 2021-07-23 | 珠海迪沃航空工程有限公司 | Dynamic image identification method and system for bolt detection |
CN113343773A (en) * | 2021-05-12 | 2021-09-03 | 上海大学 | Facial expression recognition system based on shallow convolutional neural network |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110183711A1 (en) * | 2010-01-26 | 2011-07-28 | Melzer Roy S | Method and system of creating a video sequence |
CN102271241A (en) * | 2011-09-02 | 2011-12-07 | 北京邮电大学 | Image communication method and system based on facial expression/action recognition |
-
2018
- 2018-07-25 CN CN201810830347.7A patent/CN108830262A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110183711A1 (en) * | 2010-01-26 | 2011-07-28 | Melzer Roy S | Method and system of creating a video sequence |
CN102271241A (en) * | 2011-09-02 | 2011-12-07 | 北京邮电大学 | Image communication method and system based on facial expression/action recognition |
Non-Patent Citations (2)
Title |
---|
OCTAVIO ARRIAGA ET.AL: "Real-time Convolutional Neural Networks for Emotion and Gender Classification", 《ARXIV:1710.07557V1 [CS.CV]》 * |
TONG ZHANG ET.AL: "A Deep Neural Network-Driven Feature Learning Method for Multi-view Facial Expression Recognition", 《IEEE TRANSACTIONS ON MULTIMEDIA》 * |
Cited By (38)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109784154B (en) * | 2018-12-10 | 2023-11-24 | 平安科技(深圳)有限公司 | Emotion recognition method, device, equipment and medium based on deep neural network |
CN109784154A (en) * | 2018-12-10 | 2019-05-21 | 平安科技(深圳)有限公司 | Emotion identification method, apparatus, equipment and medium based on deep neural network |
CN109800648A (en) * | 2018-12-18 | 2019-05-24 | 北京英索科技发展有限公司 | Face datection recognition methods and device based on the correction of face key point |
CN109754017B (en) * | 2019-01-09 | 2022-05-10 | 西北工业大学 | Hyperspectral image classification method based on separable three-dimensional residual error network and transfer learning |
CN109754017A (en) * | 2019-01-09 | 2019-05-14 | 西北工业大学 | Based on separable three-dimensional residual error network and transfer learning hyperspectral image classification method |
CN109829409A (en) * | 2019-01-23 | 2019-05-31 | 深兰科技(上海)有限公司 | Driver's emotional state detection method and system |
CN109948509A (en) * | 2019-03-11 | 2019-06-28 | 成都旷视金智科技有限公司 | Obj State monitoring method, device and electronic equipment |
CN110175504A (en) * | 2019-04-08 | 2019-08-27 | 杭州电子科技大学 | A kind of target detection and alignment schemes based on multitask concatenated convolutional network |
CN110222559A (en) * | 2019-04-24 | 2019-09-10 | 深圳市微纳集成电路与系统应用研究院 | Smog image detecting method and device based on convolutional neural networks |
CN110222565A (en) * | 2019-04-26 | 2019-09-10 | 合肥进毅智能技术有限公司 | A kind of method for detecting human face, device, electronic equipment and storage medium |
CN110188615A (en) * | 2019-04-30 | 2019-08-30 | 中国科学院计算技术研究所 | A kind of facial expression recognizing method, device, medium and system |
CN110188615B (en) * | 2019-04-30 | 2021-08-06 | 中国科学院计算技术研究所 | Facial expression recognition method, device, medium and system |
CN110210432A (en) * | 2019-06-06 | 2019-09-06 | 湖南大学 | A kind of face identification method based on intelligent security guard robot under the conditions of untethered |
CN110580445B (en) * | 2019-07-12 | 2023-02-07 | 西北工业大学 | Face key point detection method based on GIoU and weighted NMS improvement |
CN110580445A (en) * | 2019-07-12 | 2019-12-17 | 西北工业大学 | Face key point detection method based on GIoU and weighted NMS improvement |
CN110660046A (en) * | 2019-08-30 | 2020-01-07 | 太原科技大学 | Industrial product defect image classification method based on lightweight deep neural network |
CN110660046B (en) * | 2019-08-30 | 2022-09-30 | 太原科技大学 | Industrial product defect image classification method based on lightweight deep neural network |
CN112487855A (en) * | 2019-09-12 | 2021-03-12 | 上海齐感电子信息科技有限公司 | MTCNN (multiple-connectivity neural network) model-based face detection method and device and terminal |
CN110825852A (en) * | 2019-11-07 | 2020-02-21 | 四川长虹电器股份有限公司 | Long text-oriented semantic matching method and system |
CN110825852B (en) * | 2019-11-07 | 2022-06-14 | 四川长虹电器股份有限公司 | Long text-oriented semantic matching method and system |
CN110889672A (en) * | 2019-11-19 | 2020-03-17 | 哈尔滨理工大学 | Student card punching and class taking state detection system based on deep learning |
CN110889672B (en) * | 2019-11-19 | 2022-04-12 | 哈尔滨理工大学 | Student card punching and class taking state detection system based on deep learning |
CN111126173B (en) * | 2019-12-04 | 2023-05-26 | 玉林师范学院 | High-precision face detection method |
CN111126173A (en) * | 2019-12-04 | 2020-05-08 | 玉林师范学院 | High-precision face detection method |
CN111414884A (en) * | 2020-03-27 | 2020-07-14 | 南京工业大学 | Facial expression recognition method based on edge calculation |
CN111652146B (en) * | 2020-06-03 | 2023-03-24 | 陕西科技大学 | Detection method for facial emotional conditions of old people in nursing home |
CN111652146A (en) * | 2020-06-03 | 2020-09-11 | 陕西科技大学 | Detection method for facial emotional conditions of old people in nursing home |
CN111723709A (en) * | 2020-06-09 | 2020-09-29 | 大连海事大学 | Fly face recognition method based on deep convolutional neural network |
CN111695513B (en) * | 2020-06-12 | 2023-02-14 | 长安大学 | Facial expression recognition method based on depth residual error network |
CN111695513A (en) * | 2020-06-12 | 2020-09-22 | 长安大学 | Facial expression recognition method based on depth residual error network |
CN112304435A (en) * | 2020-10-10 | 2021-02-02 | 广州中大数字家庭工程技术研究中心有限公司 | Human body thermal imaging temperature measurement method combining face recognition |
CN112215157B (en) * | 2020-10-13 | 2021-05-25 | 北京中电兴发科技有限公司 | Multi-model fusion-based face feature dimension reduction extraction method |
CN112215157A (en) * | 2020-10-13 | 2021-01-12 | 北京中电兴发科技有限公司 | Multi-model fusion-based face feature dimension reduction extraction method |
CN112580458A (en) * | 2020-12-10 | 2021-03-30 | 中国地质大学(武汉) | Facial expression recognition method, device, equipment and storage medium |
CN112580458B (en) * | 2020-12-10 | 2023-06-20 | 中国地质大学(武汉) | Facial expression recognition method, device, equipment and storage medium |
CN113160114A (en) * | 2021-01-29 | 2021-07-23 | 珠海迪沃航空工程有限公司 | Dynamic image identification method and system for bolt detection |
CN113065460A (en) * | 2021-03-31 | 2021-07-02 | 吉林农业大学 | Establishment method of pig face facial expression recognition framework based on multitask cascade |
CN113343773A (en) * | 2021-05-12 | 2021-09-03 | 上海大学 | Facial expression recognition system based on shallow convolutional neural network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108830262A (en) | Multi-angle human face expression recognition method under natural conditions | |
Jiang et al. | A pedestrian detection method based on genetic algorithm for optimize XGBoost training parameters | |
Hemalatha et al. | A study of techniques for facial detection and expression classification | |
Yan et al. | Multi-attributes gait identification by convolutional neural networks | |
Islam et al. | Performance of SVM, CNN, and ANN with BoW, HOG, and image pixels in face recognition | |
Nguyen et al. | Deep learning for american sign language fingerspelling recognition system | |
Caroppo et al. | Comparison between deep learning models and traditional machine learning approaches for facial expression recognition in ageing adults | |
Shanta et al. | Bangla sign language detection using sift and cnn | |
Yasir et al. | Two-handed hand gesture recognition for Bangla sign language using LDA and ANN | |
Swain et al. | A review on plant leaf diseases detection and classification based on machine learning models | |
Moallem et al. | Fuzzy inference system optimized by genetic algorithm for robust face and pose detection | |
Iosifidis et al. | Neural representation and learning for multi-view human action recognition | |
CN109063626A (en) | Dynamic human face recognition methods and device | |
Chelali et al. | Face recognition using MLP and RBF neural network with Gabor and discrete wavelet transform characterization: a comparative study | |
Shanthi et al. | Algorithms for face recognition drones | |
Karayılan et al. | Sign language recognition | |
Borgalli et al. | Deep learning for facial emotion recognition using custom CNN architecture | |
Bhattacharyya et al. | Recognizing gender from human facial regions using genetic algorithm | |
Lin et al. | A new automatic recognition system of gender, age and ethnicity | |
Abd El-Rahiem et al. | An efficient deep learning model for classification of thermal face images | |
Mahmud et al. | Facial expression recognition system using extreme learning machine | |
Sun et al. | Using backpropagation neural network for face recognition with 2D+ 3D hybrid information | |
Zahid et al. | A Multi Stage Approach for Object and Face Detection using CNN | |
Chelali et al. | Face recognition system using neural network with Gabor and discrete wavelet transform parameterization | |
Vidal et al. | Dynamicboost: Boosting time series generated by dynamical systems |
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 |
Application publication date: 20181116 |
|
RJ01 | Rejection of invention patent application after publication |