search

microsoft / FocalNet

[NeurIPS 2022] Official code for "Focal Modulation Networks"
MIT License
682 stars 60 forks source link

Issue in reproducing evaluation results with sparse rcnn checkpoint #30

Open ihjasahammedm opened 1 year ago

ihjasahammedm commented 1 year ago

First of all thank you for releasing the code and ckpts. I am not able to reproduce the results using config file "sparse_rcnn_focalnet_tiny_fpn_300_proposals_crop_mstrain_480-800_3x_coco_lrf.py" and it's corresponding checkpoint given in readme. The code itself is showing that there is mismatches in state dictionary. For this to run I also had to update the line 1 in config from _base_ = '../_base_/sparse_rcnn_focalnet_fpn.py' to _base_ = '../_base_/models/sparse_rcnn_focalnet_fpn.py' image The metrics obtained are very low and close to zero! Following is the command I used to run the evaluation:

python tools/test.py configs/focalnet/sparse_rcnn_focalnet_tiny_fpn_300_proposals_crop_mstrain_480-800_3x_coco_lrf.py ckpts/focalnet_tiny_lrf_sparsercnn_3x.pth --eval bbox

Following is mAP values obtained: image Could you please look into this and let me know if I am doing something wrong here?

jwyang commented 1 year ago

thanks for your interest!

I will take a look on it.

VimukthiRandika1997 commented 2 months ago

I think in the config section there are no configs relevant to modulation based bbox_head. The one you have evaluated is attention based one. Those files are missing!