xinw1012/cycle-confusion

Code and models for ICCV2021 paper "Robust Object Detection via Instance-Level Temporal Cycle Confusion".

cycle-confusion's Issues

Which Waymo dataset should I download？

Hi, I find that there are two datasets on Waymo official website, the motion dataset and the perception dataset. And I can not find raw dataset. Which one should I download for this experiment? Thank you

Can not reach the reported performance

Hi, thanks for this excellent work!
I try to train the model using the config: R50_FPN_daytime_cycle_conf.yaml on BDD100k.
However, the performance is only 18.126 AP which is the same as my retrained baseline (18.108 AP).
There is a 1.0 AP gap with the reported performance in the paper.

Code for a conventional unsupervised domain adaptation in an image.

Thank you for your interesting work.
It was very impressive to improve the robustness to domain shift through self-supervision.

In particular, I'm interested in conventional UDA settings (e.g. cityscapes -> foggy cityscapes, Sim10k -> cityscapes) that you applied rotation operation as a self-supervision task.
Could you release the code for this, as well?