How to generate binary mask of an image through regression approach which already has half part blurred (in PYTHON)

I have one original image and from that image i have to generate the following:

The original image is: original image

1. blurred image pairs with zigzag portion (users choice) as shown below: in the first image of pair, the lower part in zigzag form is blurred and second part is exactly the opposite of this.

image a image b

2. to create binary mask of corresponding image shown in 1st point as shown below:

Mask of image a Mask of image b

Kindly tell how to generate blurred part with random portion and binary mask of the corresponding image. Also provide the code that is applied to original image and outputs the image a and b as shown in 1st point. then regression approach is applied on 1st point to generate binary mask as shown in 2nd point.

I am not able to do the above two tasks for an image. Any help will be appreciated.