Open tflahaul opened 1 year ago
The "resnet" normalization only works with Imagenet RGB mean and std: ((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)) Being able to specify the mean & std of other datasets at instantiation could be great.
((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
Current workaround:
f = aloscene.Frame(torch.rand(1, 10, 10), names=('C', 'H', 'W'), normalization="resnet") f.mean_std = ((0.5, 0.5, 0.5), (0.3, 0.3, 0.3))
Also, changing the name of the "resnet" normalization to "z-norm", "meanstd" or something similar could be better for everyones understanding
The "resnet" normalization only works with Imagenet RGB mean and std:
((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))Being able to specify the mean & std of other datasets at instantiation could be great.Current workaround:
Also, changing the name of the "resnet" normalization to "z-norm", "meanstd" or something similar could be better for everyones understanding