perceivelab / hd2s Goto Github PK

Hello,
We were trying your test.py code and initially found the heatmap output quite strange. After reading the HD2S.py code, we think the reason is that in this line x was returned instead of out. Because x is the logit, it can be negative. On the other hand, this line apparently assumes the output of the model is positive. So it will produce strange heatmaps. Changing the model to return out fixed the problem.

Hi, can you provide a license for this repo?

hi, i have some questions about the results on Hollywood2 and UCFsports.

1. The result for hd2s on dataset hollywood2 is train on dataset DHF1K's trainset, and test on dataset hollywood2's testset, right?
   Because i view that the result of hd2s for one-source learning is different from the above result.
   
   

2.Why you use DHF1K Hollywood2 and UCFsports as validation set in train_DSL.py? And there aren't splitTrainVal floder for Hollywood2.

But i view that in the paper:

Looking forward to hearing from you.Thank you so much！