Few-sample behavior recognition method based on meta-learning
A recognition method and meta-learning technology, applied in the direction of neural learning methods, character and pattern recognition, instruments, etc., can solve problems such as people's difficulties, and achieve the effect of good generalization ability between tasks
Pending Publication Date: 2021-06-29
ZHEJIANG UNIV OF TECH
View PDF0 Cites 2 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Although the technical progress of deep learning in the past ten years has greatly improved the accuracy of video behavior recognition, labeling such a large amount of video data has brought great difficulties to people.
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0039] The present invention will be further described below in conjunction with the accompanying drawings.
[0040] refer to figure 1 and figure 2 , a few-shot behavior recognition method based on meta-learning, including the following steps:
[0041] 1) Divide the video dataset into a meta-training set and a meta-testing set, extract multiple sets of support sets and query sets from the meta-training set for training the model, and extract multiple sets of support sets and query sets from the meta-testing set for testing the model.
[0042] Divide the video dataset into a meta-training set D meta-train and the meta-test set D meta-test , during training, each round starts from D meta-train Randomly select N different classes, and each class has K different samples to form a support set Then from the remaining D meta-train Randomly select samples from these N classes to form a query set During the test, the D meta-test do the same.
[0043] 2) Using a shallow 3D c...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
A few-sample behavior recognition method based on meta-learning comprises the following steps: 1) dividing a video data set into a meta-training set and a meta-testing set, extracting multiple groups of support sets and inquiry sets from the meta-training set for training a model, and extracting multiple groups of support sets and inquiry sets from the meta-testing set for testing the model; 2) extracting video features of a support set and an inquiry set by using a shallow three-dimensional convolutional neural network; 3) constructing a meta-learning network for modeling the support set to generate parameters of the three-dimensional convolutional neural network of the shallow layer in the step 2); 4) performing second-order transformation and normalization processing on the video features extracted in the step 2); and 5) splicing the processed video features of the support set and the inquiry set, extracting a nonlinear distance relationship between the video features of the inquiry set and the support set by using a multilayer two-dimensional convolutional neural network, and classifying the videos of the training set. The method has good inter-task generalization ability and recognition accuracy of new video behaviors.
Description
technical field [0001] The invention relates to the technical field of video behavior recognition, in particular to a few-sample behavior recognition method based on meta-learning. Background technique [0002] Behavior recognition technology is one of the research focuses in the field of computer vision, and it is widely used in urban traffic control, smart security and other fields. [0003] With the rapid development of network technology and the large-scale installation of smart cameras, video data is growing explosively every day. Although the technical progress of deep learning in the past ten years has greatly improved the accuracy of video behavior recognition, labeling such a large amount of video data has brought great difficulties to people. Not only that, it is still relatively rare to collect videos in specific fields, such as abnormal behavior scenes, dangerous behaviors in factories, etc. How to use only a small amount of sample data to train the model and o...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More IPC IPC(8): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V20/41G06N3/045G06F18/22G06F18/214
Inventor 陈朋宗鹏程党源杰俞天纬王海霞
Owner ZHEJIANG UNIV OF TECH