Improve projective data augmentation

[lorenzoh]

Currently augmentation on images and keypoints is restricted to a random crop by default. fastai uses much more aggressive augmentation, and so should we. This should be straightforward to implement using DataAugmentation.jl's composable ProjectiveTransform.

The result should be something mirroring the functionality of fastai's aug_transforms.

To accomplish this, the following affine transformations should be added to DataAugmentation.jl:

• Vertical and horizontal flip
• Random rotation
• Random warp
• Random lighting
• Random zoom

And then a helper in FastAI.jl that creates the augmentations which can be passed to ProjectiveTransforms.

[lorenzoh]

These augmentations will be added to DataAugmentation.jl first and then used in FastAI.jl.

• Random lighting (brightness and contrast adjustment) DataAugmentation.jl#28
• Warp DataAugmentation.jl#30
• Zoom DataAugmentation.jl#31

[lorenzoh]

Implemented with the DataAugmentation.jl 0.2.2 release and FastAI.augs_projection, FastAI.augs_color. See this tutorial.