i can't find where you use finetuning on first frame in your code-pytorch about osvos-pytorch

[kmaninis]

Hi, online training on the first frame happens in this script.

The data loader uses only the first frame in this part of the dataloader.

I hope it helps.

[zouying-sjtu]

one more question, besides the dataset is different, is there any other difference between training and finetuning. thanks

[kmaninis]

Could you be more specific?

[zouying-sjtu]

when you train the model, you use many data-label pairs from trainset. when you finetune, you just use the pair of the first frame. is there other more manipulation you do?

[kmaninis]

The offline and online training processes are almost identical. Some details are different, eg. the deeply supervised loss in offline training, and parameters such as nAveGrad, since more speed is needed during online training.