haochenheheda / LVVIS

Large-Vocabulary Video Instance Segmentation dataset
GNU General Public License v3.0
76 stars 1 forks source link

Reproduce OV2Seg results on LVVIS val and test set #11

Open sumcher-ng opened 1 year ago

sumcher-ng commented 1 year ago

Thank you for your great work.

I used your scripts/train.sh to train ov2seg with Resnet50 backbone on 4 gpus and batch size 8 with total 370k iterations (all other hparams are kept unchanged), but cannot obtain the reported numbers in table 2 in the paper.

Do I need to config anything else? or can you please share the training log of the provided R50 version? Thank you!

haochenheheda commented 12 months ago

Sorry that there is a mistake in the paper. We checked the log and the batch size was 16 indeed. Please try it with bs=16. We will fix the typo in the paper.