You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi everyone,
I implemented a pytorch version of TSN, and the performence of my TSN on UCF101 is perfectly fine.
However, according to the original paper (https://arxiv.org/pdf/1705.02953.pdf), the author mentioned that TSN gets mAP 79.4% on the validation set of ActivityNet v1.3. I strictly follow the training settings in the paper to train my TSN on ActivityNet v1.3 and the testing methods (M-TWI) proposed in the paper, but I can only get mAP around 73%-74%.
Apparently, there must be something wrong in my implement. Did anyone reproduce TSN on ActivityNet v1.3? Or are there any pretrained models on ActivityNet v1.3 available? Any clues would be greatly appreciated. Thanks!
The text was updated successfully, but these errors were encountered:
Hi everyone,
I implemented a pytorch version of TSN, and the performence of my TSN on UCF101 is perfectly fine.
However, according to the original paper (https://arxiv.org/pdf/1705.02953.pdf), the author mentioned that TSN gets mAP 79.4% on the validation set of ActivityNet v1.3. I strictly follow the training settings in the paper to train my TSN on ActivityNet v1.3 and the testing methods (M-TWI) proposed in the paper, but I can only get mAP around 73%-74%.
Apparently, there must be something wrong in my implement. Did anyone reproduce TSN on ActivityNet v1.3? Or are there any pretrained models on ActivityNet v1.3 available? Any clues would be greatly appreciated. Thanks!
The text was updated successfully, but these errors were encountered: