WeianMao

Thanks for instruction. However, as my understanding, resume_from needs to manully set the path of the checkpoint file. In my case, a job may needs to resume by 30 times....

I modify some code. this issue has been addressed. add the code in train.py work_dir = cfg._filename.split('/')[-1].split('.')[0] work_dir = 'work_dirs/' + work_dir + '/' if 'latest.pth' in os.listdir(work_dir): cfg.resume_from =...

yes, supervise on larger feat can boost the performance

what is your batch size?

> Hi @ssssholmes, > > 1. We use the original optimizer (sgd) and learning rate in Mask R-CNN, with 180000 iters. And the. Mask R-CNN model is pre-trained with heatmap....

great! thank you!

wonderful! thank you very much!

同问

Hi, all the models weight and cig file have been released, pls check read me. However, due to the recent update of mmpose, some model may have worse performance when...

yes, our ori repo is based on mmpose. But i think the current mmpose modify some code. that causes some diffence, so the result of the current repo may differrent...