Can you provide the performance based on the GT boxes? about grit HOT 3 CLOSED

Hi, @Weiyun1025!

I've checked the results in All-Seeing, and its performance is quite good. Would you mind sharing the code of promptable GRiT?

Thanks in advance!

I've released an app for promptable GRiT https://github.com/xk-huang/Promptable-GRiT.

from grit.

Hi, @Weiyun1025!

I've checked the results in All-Seeing, and its performance is quite good. Would you mind sharing the code of promptable GRiT?

Thanks in advance!

from grit.

Hi, @Weiyun1025!
I've checked the results in All-Seeing, and its performance is quite good. Would you mind sharing the code of promptable GRiT?
Thanks in advance!

I've released an app for promptable GRiT https://github.com/xk-huang/Promptable-GRiT.

Thank you for your interest in our work and for sharing the code!

To assess the performance of the GRiT model, we directly modified the code within this repository, so that we can input the ground truth boxes to the model and skip the post-processing steps for these boxes. As our objective is evaluation rather than deployment, we do not currently have a polished version of the code that is ready for release.

We believe that your code will be highly beneficial for others who are also interested in the GRiT model!

from grit.