search

foolwood / DCFNet

DCFNet: Discriminant Correlation Filters Network for Visual Tracking
MIT License
214 stars 66 forks source link

Does VID training use 86G data? #11

Closed xiaoxiaohahahaha closed 6 years ago

xiaoxiaohahahaha commented 6 years ago

Hello, does the training data use VID's latest 86G training? Why do I get out of memory errors when running train_DCFNet.m?

foolwood commented 6 years ago

We use the 86G training data in http://bvisionweb1.cs.unc.edu/ilsvrc2015/download-videos-3j16.php#vid

We put all the crop in memory. (About 20G RAM) to speed up the training process.

If your PC ram is 8G, I suggest that you can set a small value in https://github.com/foolwood/DCFNet/blob/97d2cd784d9c2b1083c1249a2aef914062fb5910/training/getImdbDCFNet.m#L21

and use a subset of VID (n_videos =1000).

https://github.com/foolwood/DCFNet/blob/97d2cd784d9c2b1083c1249a2aef914062fb5910/training/getImdbDCFNet.m#L41

I will fix this problem on April.

foolwood commented 6 years ago

You can try the PyTorch implementation.
https://github.com/foolwood/DCFNet_pytorch

It's much faster and use less RAM. You can train it on a PC with 8G RAM and a cheap GTX1060 in one day.