Added patch for folding augmentation.

[kwcckw]

Basically gradient width and height of folding is based on % of image width and height. If the image size is too small , the width or height range will be the same and causing those errors. We would need 4 unique points for the perspective transform. I added an if loop to show warning message and prevent further computation if there is no 4 unique points.

[proofconstruction]

I tested this and still found errors (see below), but I'm still merging this.

Bug

fold_width_one_side can be 0 (line 193) if images are less than 20x20. I didn't see this originally when I found the bug with 10x10 images. If images are larger than 20x20, this doesn't seem to happen, so I won't consider this a problem. We will deal with minimum image sizes some other way, see discussion in #58

[kwcckw]

I tested this and still found errors (see below), but I'm still merging this.

Bug

fold_width_one_side can be 0 (line 193) if images are less than 20x20. I didn't see this originally when I found the bug with 10x10 images. If images are larger than 20x20, this doesn't seem to happen, so I won't consider this a problem. We will deal with minimum image sizes some other way, see discussion in #58

Thanks, you are right. So a better way is to check the folding width:

if (fold_width_one_side != 0) and (fold_y_shift !=0) :

I will include this in the next update: