A transformer-based multi-view target detection method and system
A target detection, multi-view technology, applied in the field of target detection, can solve problems such as imperfect occlusion problems, and achieve the effect of avoiding poor performance, improving accuracy, and reducing model calculation time
Active Publication Date: 2022-03-15
TSINGHUA UNIV
View PDF7 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0003] However, these methods often contain a lot of redundant calculations, and the solution to the occlusion problem is not perfect.
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
[0048] In order to make the purpose, technical solution and advantages of the application clearer, the technical solution of the application will be clearly and completely described below in conjunction with specific embodiments of the application and corresponding drawings. It should be understood that the described embodiments are only some of the embodiments of the present application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the scope of protection of this application.
[0049] Before introducing the embodiments of the present invention, the relevant terms involved in the embodiments of the present invention are first explained as follows:
[0050] Multi-view camera: refers to multiple monocular cameras placed at the intersection, distributed on the roadside, and the total field of view of the multi-view camera can cover the enti...
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
The present invention provides a method and system for multi-view target detection based on Transformer. The method includes: using multiple cameras to simultaneously collect RGB images of multiple viewing angles and preprocessing; Input the trained multi-view target detection model, and output a bird's-eye view including the target detection result; the multi-view target detection model includes: a feature extraction module, a Transformer model and a projection module; the feature extraction module is used to extract each The multi-scale feature map of the RGB image of the perspective, the multi-scale feature map of multiple perspectives is input into the Transformer model; the Transformer model is used to perform target detection on the input feature map, and output the bounding box; the projection module is used for A Gaussian heatmap is generated centered on the midpoint of the bounding box predicted by the Transformer model, fused with the feature maps from multiple perspectives output by the feature extraction module, and then a bird's-eye view is output after projection transformation and convolution.
Description
technical field [0001] The invention relates to the field of target detection, in particular to a Transformer-based multi-view target detection method and system. Background technique [0002] For pedestrian detection under occlusion, the existing related work is summarized, and the occlusion problem is mainly solved through the angle of single-view detection. The entry point of the current single-view detection method is to divide the target candidate frame into different parts and process them one by one, treat them differently and then add feature fusion, or make the target candidate frame more discriminative for mutual occlusion from the perspective of loss. By setting the loss function In this way, the distance between the predicted frame and the responsible real target frame is reduced, and the distance between it and the surrounding non-responsible target frame (including the real target frame and the predicted frame) is increased to improve the model performance. ...
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 Patent Type & Authority Patents(China)
IPC IPC(8): G06V40/10G06V40/20G06V10/25G06V10/80G06V10/82G06K9/62G06N3/04
CPCG06N3/045G06F18/253
Inventor 张新钰李志伟李骏高鑫刘宇红王力杜浩
Owner TSINGHUA UNIV