search

zju3dv / disprcnn

Code release for Stereo 3D Object Detection via Shape Prior Guided Instance Disparity Estimation (CVPR 2020, TPAMI 2021)
Apache License 2.0
213 stars 36 forks source link

About segmentation mask #9

Closed dropletpeace closed 4 years ago

dropletpeace commented 4 years ago

Thanks for sharing your code with us. I have a question about how you trained your mask-rcnn since there is no GT mask. Looking forward to your response. Best wishes~

JiamingSuen commented 4 years ago

The mask head is trained with the generated instance masks pseudo-gt rendered together with the instance disparity pseudo-gt. If you need more detailed information like where those pseudo-gt masks are located, @f-sky can provide more information.

Additionally, manually annotated dataset like KINS would also be useful if you need more accurate mask boundaries.

Cheers~

JiamingSuen commented 4 years ago

Feel free to reopen this issue if you have more question.