search

TencentYoutuResearch / CrowdCounting-P2PNet

The official codes for the ICCV2021 Oral presentation "Rethinking Counting and Localization in Crowds: A Purely Point-Based Framework"
Other
423 stars 114 forks source link

the input image size #64

Open ycbilge opened 1 year ago

ycbilge commented 1 year ago

Is there any fixed image size for the model? For different image sizes I get different results for the shangaiTech dataset. It is especially an issue when working with the onnx model.