Image normalization

Open mengdaxu opened this issue 3 years ago • 2 comments

Thanks for sharing the pre-train model! I noticed that you have normalized the image input by im_mean and im_std in the simulation environment. To use the pre-train model, are we expected to normalize the image input ( from real-world or another simulation env) by the same mean and std?

Thanks!

Oct 25 '22 03:10 mengdaxu

Bump. I'm getting bad results with a real robot and I'm wondering if normalization has to do with it.

Jun 13 '23 07:06 seann999

Bump again. I'm using ImageNet constants with mean [0.485, 0.456, 0.406] and std [0.229, 0.224, 0.225] but can't get decent results. It would be helpful if the example in the readme contained lines of code for inputting images.

Oct 13 '23 07:10 seann999