Sauradip Nag
Sauradip Nag
Thank you for the interest in our work ! I am currently writing my thesis, so it's difficult to make time but expect the code release in the last week...
Thanks all for the interest. November Last Week the code should be uploaded. Apologies for the delay !
Hi all thisbwork happened during my PhD ending phase and since i graduated i lost access to the system to test it properly. This is the reason i could not...
We are trying to release the code before cvpr deadline. Since the implementation is based on GTAD , in our results it showed poor performance as seen from the main...
Hi , it is easy to train , hardly takes 15 - 30 mins to train and get the checkpoint. I lost the system as i graduated as PhD. Kindly...
Hi, Although the proposals are GT, the class prediction will be different. So when you combine GT proposals with predicted Class it will be different
A1. It is true , this approach will lead to over generalization of long action instances. We need to change the inference script and not include the filling operation. This...
A1. I did not use filling for THUMOS. A3-1. Both can be used, yes for ActivityNet, it comes out more consistent than maskformer ( if i use 1 query )....
Hi, Your 2D action mask (250x250) looks not that bad, I can see initial two masks at top left , some in the middle and then some in bottom right...
Hi , I will recommend you to check with the UNet value as a class score refinement just like it is done here for activitynet postprocessing since it is a...