search

jfzhang95 / pytorch-video-recognition

PyTorch implemented C3D, R3D, R2Plus1D models for video activity recognition.
MIT License
1.18k stars 250 forks source link

The accuracy of C3D training from scartch is low #32

Closed Fancy-sf closed 4 years ago

Fancy-sf commented 5 years ago

The accuracy of C3D training from scratch is 30% with lr=1e-5 and the accuracy is below 1% with lr=1e-3 , which are lower than the paper claimed. I also notice someone added BN to C3D and the accuracy is about 45%. Anybody knows why?

civat commented 5 years ago

The acc reported in this repo may be not based on the implementation in this repo. The performance can simply improved by standardizing the input:

Normalize input into [0, 1] by x/255; Standardize input using mean [0.485, 0.456, 0.406] and variance [0.229, 0.224, 0.225].

After standardizing, the acc is about 50%.

skyqwe123 commented 3 years ago

@civat What about R2Plus1D? The accuracy of R2Plus1Dtraining from scratch is also very lower (1%)