search

amazon-science / long-short-term-transformer

[NeurIPS 2021 Spotlight] Official implementation of Long Short-Term Transformer for Online Action Detection
Apache License 2.0
127 stars 19 forks source link

BN-Inception model recommended #26

Open dqj5182 opened 1 year ago

dqj5182 commented 1 year ago

In the readme of this GitHub repo, it says "For Kinetics pretrained features, we use the ResNet-50 model for the RGB inputs. We recommend to use this checkpoint in MMAction2. We use the BN-Inception model for the optical flow inputs. We recommend to use the model here".

The link for "here" is not working. Could you fix it?

Prism-hua commented 6 months ago

Hello @dqj5182 , I have the same problem with you, I am writing to ask whether you have solved that. If yes, could you please tell me where can I get that file? Thanks.

Hello @xumingze0308, congratulations for your great work, I am having difficulties extracting features of the TVSeries datasets. The checkpoint and file mentioned in README do not work now. Could you please fix that or share me with other solutions? Thanks so much.

dqj5182 commented 6 months ago

I just used TESTRA's pretrained checkpoint for this! Because I did not need exact reproducibility, this worked for me.

This is the link for TESTRA, the future work of LSTR: https://github.com/zhaoyue-zephyrus/TeSTra