ColorJitter Enhancement

[yaox12]

Set custom min and max bounds by inputing a tuple for each parameter in ColorJitter.

See issue #545

[fmassa]

Thanks for the PR!

I've left a few comments that I think could improve the functionality and make it less error-prone if in the future we want to extend its functionality.

[yaox12]

I've improved it according to your advice. Please review it. @fmassa

[vfdev-5]

@yaox12 just wonder in which case the following conditions are met like

if brightness_factor is not None:

in _sample_from I think there is no possibility to return None...

[fmassa]

I made another round of comments. Let me know what you think

[yaox12]

In the previous code, if brightness = 0, then nothing would happen. In my code, it will call F.adjust_brightness(img, 1), though it will return img without any change.
I would like to make this case return None, but I find it will bring more if. Do you have any elegant solution? @vfdev-5

[vfdev-5]

@yaox12 Maybe something like in the previous code could help. Previously, if brightness/saturation/... value is 0 than ignore it, we can replace it with None and check as previously :

if saturation is not None:
    saturation_factor = _sample_from(saturation, 'saturation')
    transforms.append(Lambda(lambda img: F.adjust_saturation(img, saturation_factor)))

or if we want to keep it as before

if saturation > 0:
    saturation_factor = _sample_from(saturation, 'saturation')
    transforms.append(Lambda(lambda img: F.adjust_saturation(img, saturation_factor)))

HTH

[yaox12]

Check input in __init__. Welcome for comments.

[fmassa]

LGTM, thanks!