Human body part analyzing method and system, equipment and storage medium
An analysis method and human body technology, applied in the field of computer vision, can solve problems such as low analysis accuracy, and achieve the effect of reducing the difficulty of analysis, improving the analysis accuracy, and reducing randomness
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0079] Such as figure 1 As shown, the human body part parsing method based on multi-person images of this implementation includes:
[0080] S101. Extracting a first feature map with high-level semantic information from a multi-person image;
[0081] Among them, the high-level semantic information is used to represent the overall information in the multi-person image, such as the objects in the image, the ongoing actions of the objects, and the overall scene information.
[0082] S102. Obtain a plurality of first human body regions of interest according to the first feature map;
[0083] S103. For each first human body region of interest, select a target human body object from the first human body region of interest, and expand the first human body region of interest into a second human body region of interest;
[0084] Wherein, the relatively fixed position of each target human body object in the corresponding second human body interest region makes the target human body obj...
Embodiment 2
[0091] Such as figure 2 and image 3 As shown, this embodiment is further improved on the basis of embodiment 1, specifically:
[0092] Step S101 specifically includes:
[0093] S1011. Use Deeplab v2 (a model for image semantic segmentation formed by using a deep convolutional network) to obtain a first feature map with high-level semantic information in a multi-person image. Specifically, based on the first five convolutional layers of the deep convolutional network, the first feature map is obtained;
[0094] The high-level semantic information includes at least one of color features, texture features, shape features and spatial relationship features in the image.
[0095] Step S102 specifically includes:
[0096] S1021. According to the first feature map, use RPN (Region Proposal Network, region proposal network) to acquire a first human body region of interest.
[0097] Specifically, the principle of using the region proposal network to obtain the region of interest ...
Embodiment 3
[0152] Such as Figure 4 As shown, the system of human body parts analysis based on multi-person images in this embodiment includes a first feature map acquisition module 1, a first area acquisition module 2, a second area acquisition module 3, a second feature map acquisition module 4 and a first feature map acquisition module. Analysis module 5.
[0153] The first feature map acquisition module 1 is used to extract the first feature map with high-level semantic information from the multi-person image;
[0154] Among them, the high-level semantic information is used to represent the overall information in the multi-person image, such as the objects in the image, the ongoing actions of the objects, and the overall scene information. The first region obtaining module 2 is used to obtain a plurality of first human body regions of interest according to the first feature map;
[0155] The second region acquiring module 3 is used for selecting a target human body object from the ...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com