alvin-zeng / drn Goto Github PK

RuntimeError: CUDA out of memory. Tried to allocate 2.00 MiB (GPU 0; 10.92 GiB total capacity; 9.70 GiB already allocated; 3.00 MiB free; 10.39 GiB reserved in total by PyTorch)
How could I fix this? I had changed the batchsize to 2 but it still come to the error.
It there anything special in second stage?

Hi, I'm interested in your work. I have already download C3D features. Could you provide your extracted I3D features of Charades, ActivityNet and TACoS dataset? Thank you very much.

Hi, when I run train.py, there always exists this error:
ImportError: cannot import name '_C' from 'fcos_core' (/home/PycharmProjects/DRN-master/fcos_core/__init__.py)
could you help me?

Hello, thank you for sharing the code!

From the DRN paper, I know DRN follows anchor-free methods and achieves great performance on 3 temporal grounding datasets. By the similar architecture, I believe that DRN can also achieve a good result in temporal action localization task, but there is no R@1 IoU=0.5 metric in temporal action localization task. I want to know whether you mean mAP IoU=0.5 or the metric like AR@100 in temporal action proposal task?

P.S. It will be better if you can share the code for action localization/proposal tasks in THUMOS14 dataset.

Thank you again for sharing the code of this inspiring work!

Hi Alvin, your work is remarkable! Do you have any plan to release the source code?

Hello,

Can you share the code and feature data of TACoS and ActivityNet-Captions datasets? Although the paper reports performance on all 3 datasets, this repo only contains code and data of Charades-STA.

Thank you very much!