Unlock instant, AI-driven research and patent intelligence for your innovation.

Video behavior recognition method based on improved MobileNet

A recognition method and behavior technology, applied in character and pattern recognition, instruments, biological neural network models, etc., to achieve the effect of high recognition accuracy

Pending Publication Date: 2022-07-08
CHINA JILIANG UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, most of these models are designed for simple action classification on static images, and cannot be used for complex action recognition and evaluation with strong temporal relationships.

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 more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Video behavior recognition method based on improved MobileNet
  • Video behavior recognition method based on improved MobileNet
  • Video behavior recognition method based on improved MobileNet

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0057] Example 1: Behavior Recognition Test in UCF101 Database

[0058] UCF101 is an action recognition dataset for realistic action videos, collected from YouTube, providing 13320 videos from 101 action categories. Videos in 101 action categories are divided into 25 groups, each group can contain 4-7 videos of one action. UCF101 offers the greatest diversity in action, with large variations in camera motion, object appearance and pose, object scale, viewpoint, cluttered backgrounds, lighting conditions, and more. The behavior recognition results of this embodiment are compared with the methods of VGG16, VGG19, ResNet101, ResNet152, ShuffleNet, SqueezeNet, and MobileNet. Table 1 shows the comparative experimental results of the method proposed by the present invention and seven existing methods. As can be seen from Table 1, compared with the existing behavior recognition methods, the method of the present invention has obvious advantages in terms of accuracy (precise rate) a...

Embodiment 2

[0061] Example 2: Behavior Recognition Test on HMDB51 Dataset

[0062] HMDB51 contains 51 categories of actions, a total of 6849 videos, each action contains at least 51 videos, and the sample resolution is 320*240. Each annotation is verified by at least two humans to ensure consistency. The recognition results of this embodiment are compared with the methods of VGG16, VGG19, ResNet101, ResNet152, ShuffleNet, SqueezeNet, and MobileNet. Table 2 shows the comparative experimental results of the method proposed by the present invention and seven existing methods. As can be seen from Table 2, compared with the existing behavior recognition scheme, the method of the present invention has significant advantages in the overall prediction accuracy and the like.

[0063] Table 2 Comparative test results on the HMDB51 dataset

[0064] VGG16 VGG19 ResNet101 ResNet152 ShuffleNet SqueezeNet MobileNet method of the invention Accuracy 0.68 0.71 0.85 0.8...

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

No PUM Login to View More

Abstract

The invention provides an improved MobileNet-based video behavior recognition method, which comprises the following steps of: firstly, inputting a plurality of continuous behavior frames in a behavior video to be recognized into a proposed recognition network to mine motion trend characteristics between front and back frames in the behavior video; and then, in a weighted point-by-point convolution process, a random fade-in factor is added on a time axis, and different weights are provided for each related frame to more effectively utilize a motion trend relationship between behavior frames at different moments. According to the behavior recognition method provided by the invention, a lightweight network architecture based on multi-frame MobileNet is adopted, and a plurality of continuous behavior frames are introduced to describe internal differences of similar behaviors, so that fine-grained behavior recognition, detection and evaluation are realized, and the behavior recognition accuracy based on videos is improved.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, and more particularly, to a video behavior recognition method based on an improved MobileNet. Background technique [0002] Video behavior recognition is an important research direction in the field of computer vision, and has great application potential in intelligent video surveillance, motor behavior assessment, gait recognition, etc. Simple video action recognition, i.e. action classification, only needs to correctly classify a given video into several known action categories, while complex action recognition usually consists of a set of actions with a strict temporal order relationship, rather than only a set of actions in a video. An action category. In addition, video content and backgrounds are more complex and varied than static images. Different behaviors may have similarities, and the same behavior can behave differently in different contexts. [0003] Thanks to the ...

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
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06V40/20G06V20/40G06V10/764G06V10/80G06V10/82G06N3/04G06K9/62
CPCG06N3/047G06N3/045G06F18/2415G06F18/253Y02D10/00
Inventor 王修晖刘琳琦王亚茹李学盛贾波包其富
Owner CHINA JILIANG UNIV