search

zhiyuanyou / SAFECount

[WACV 2023] Few-shot Object Counting with Similarity-Aware Feature Enhancement
Other
123 stars 22 forks source link

Visualization & post processing #4

Closed jaideep11061982 closed 1 year ago

jaideep11061982 commented 1 year ago

@zhiyuanyou I am viewing the below density dotted image. How can we interpret which heatmapped dot are likely to be counted. Currently I do simply output.sum() ,is there any way i can do sum based on some threshold value which indicates high likelyhood of an object count ?

image

zhiyuanyou commented 1 year ago

Actually, you could not find the accurate localization of the counted objects, since this method is density-based. If you wanna count objects by accurately localizing objects, you should seek to some point-based methods.

jaideep11061982 commented 1 year ago

@zhiyuanyou do you have any pointers, but density labels that were chose for the models training were point based only

zhiyuanyou commented 1 year ago

If you wanna find the accurate localization of the counted objects, I recommend CrowdCounting-P2PNet for crowd counting and Counting-DETR for few-shot counting.

If you try to reproduce P2PNet, please pay attention to the following bug (which is I find during my reproduction): Bug.