Are the codes for TTA available?

[Cascol-Chen]

It would be of great help, please let me know if you can share the code

[suniique]

Hey @Cascol-SCUT, thanks for the question!

Yes, we do have a repo working with the TTA on the SHIFT dataset. Feel free to check it out at https://github.com/zwbx/SHIFT-Continuous_Test_Time_Adaptation. You can ask for more information there!

Also, we are organizing challenges based on SHIFT in the VCL Workshop at ICCV2023. The TTA is also part of the challenges; welcome to participate!

[mattiasegu]

Hi @Cascol-SCUT, we also provide a codebase for test-time adaptation of object detection.

You can find it here: https://github.com/SysCV/shift-detection-tta

[Cascol-Chen]

Hey @Cascol-SCUT, thanks for the question!

Yes, we do have a repo working with the TTA on the SHIFT dataset. Feel free to check it out at https://github.com/zwbx/SHIFT-Continuous_Test_Time_Adaptation. You can ask for more information there!

Also, we are organizing challenges based on SHIFT in the VCL Workshop at ICCV2023. The TTA is also part of the challenges; welcome to participate!

Thanks for your kind reply. Interestingly, I find that Tent does not improve performance compared with source model, which contradicts to the paper in Figure 6. Could you further explain the difference?