search

mmaaz60 / mvits_for_class_agnostic_od

[ECCV'22] Official repository of paper titled "Class-agnostic Object Detection with Multi-modal Transformer".
MIT License
299 stars 25 forks source link

Loading checkpoints for inference #14

Closed KaleemW closed 2 years ago

KaleemW commented 2 years ago

Which checkpoints in drive link you provided will load correctly in default MDefDETR model without any errors? Im getting missing/unexpected keys errors.

mmaaz60 commented 2 years ago

Hi @KaleemW,

Please try loading MDef_DETR_r101_epoch20.pth checkpoints. If you will still get an error then please post the full error message along with the process you are following for loading the weights here. I hope it will be helpful. Thank you

KaleemW commented 2 years ago

Thanks a lot. It worked. Just wanted to know one last thing that is there a cpu implementation available for this project. I ran successfully with cuda but without cuda it doesnt compile MSDA...

mmaaz60 commented 2 years ago

Thanks a lot. It worked. Just wanted to know one last thing that is there a cpu implementation available for this project. I ran successfully with cuda but without cuda it doesnt compile MSDA...

Hi @KaleemW,

Unfortunately, only GPU implementation is available at the moment.

KaleemW commented 2 years ago

Ok Sir! Thank You for your help.

Edwardmark commented 1 year ago

@mmaaz60 hi could you tell me what is the inference speed of your model? and how to do batch inference to improve the efficiency? Thanks.

mmaaz60 commented 1 year ago

Hi @Edwardmark,

Thank you for your interest in our work. Please refer to the Table 10 of our paper for inference speed. Further, the batch inference implementation is available at modulated_detection.py.

I hope it will help. Thank You.