Questions about SCUT-Head detection model reproduction and settings

[eunseopski]

Hi, thank you for sharing your great work.

I have a few questions about reproducing the detection model results shown in your paper:

1. Is there a pretrained detection model (weights) trained on SCUT-Head dataset? (I'm referring to the model in Table 2 of your paper.)

2. In the codebase, there is a config file ScutHead_R50.yaml under the SCUT config folder. How was this config file used?

3. The pretrained model you provided uses upscale_rpn = False, and the anchors for each feature map are matched like comb_anchors = ((12,), (32,), (64,), (112,)). This seems different from what is described in the paper. Could you explain why?

Thank you very much for your help!