Closed Cascol-Chen closed 1 year ago
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!
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
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?
It would be of great help, please let me know if you can share the code