Overall framework of point-based crowd counting with contrastive learning.

<p>The blue dashed box includes four different-sized feature layers from the VGG backbone network. The backbone network section can be replaced with other structures such as ResNet. The dashed box containing the projection head, contrastive loss, and point matching is used only during the trai...

Full description

Saved in:

Bibliographic Details
Main Author:	Rundong Cao (18449715) (author)
Other Authors:	Jiazhong Yu (21647712) (author), Ziwei Liu (1877539) (author), Qinghua Liang (741665) (author)
Published:	2025
Subjects:	Cell Biology Infectious Diseases Space Science Biological Sciences not elsewhere classified two key challenges quality feature maps numerous missed detections model &# 8217 current methods struggle convolutional neural network comparative experimental results world monitoring scenarios scale target groups predict head points crowd counting based crowd counting scale targets xlink "> variable backgrounds training data towards real supervised learning regression loss paper proposes open environments obtain high labeled data image patch foreground heads different scales detecting targets complex scenarios
Tags:	Add Tag No Tags, Be the first to tag this record!

Overall framework of point-based crowd counting with contrastive learning.

Similar Items