search

ohhhyeahhh / SiamCAR

SiamCAR: Siamese Fully Convolutional Classification and Regression for Visual Tracking (CVPR 2020, Oral)
https://openaccess.thecvf.com/content_CVPR_2020/html/Guo_SiamCAR_Siamese_Fully_Convolutional_Classification_and_Regression_for_Visual_Tracking_CVPR_2020_paper.html
323 stars 63 forks source link

Reproduce the result #61

Open Zongwei97 opened 3 years ago

Zongwei97 commented 3 years ago

Thank you for the brilliant job.

I am reproducing the result with GOT10k. With the checkpoint epoch 19, I cannot get the same result as presented in the paper (huge gap).

Should I test with each saved checkpoint from 10 to 19 ?

Thanks for your help

ohhhyeahhh commented 3 years ago

Thank you for the brilliant job.

I am reproducing the result with GOT10k. With the checkpoint epoch 19, I cannot get the same result as presented in the paper (huge gap).

Should I test with each saved checkpoint from 10 to 19 ?

Thanks for your help

Hi, All the results in our paper are test with checkpoint epoch 20.

BTW, Our new model got much higher precision result, welcome to try: https://github.com/ohhhyeahhh/SiamGAT

huchenjie339 commented 3 years ago

@ohhhyeahhh 你好，最近在跑Siamcar的时候，训练了20个poch，log文件显示了训练了20个poch，但最终只保存了19个.pth（1-19）文件，为了能够保存第20个.pth文件，我在if epoch == cfg.TRAIN.EPOCH:下面加了 torch.save( {'epoch': epoch, 'state_dict': model.module.state_dict(), 'optimizer': optimizer.state_dict()}, cfg.TRAIN.SNAPSHOT_DIR+'/checkpoint_e%d.pth' % (epoch)) 但最终也只保存了19个.pth。这是什么呢？希望作者能帮忙解答下

huchenjie339 commented 3 years ago

@WangJun-ZJUT