Patching subimages into an image using vectorization

vision
1

Hey, I have this following piece of code:

# image is a torch tensors being on the gpu representing an image
# bounding boxes is a list of lists describing 4 coordinates

for bb in bounding_boxes:

    x1 = bb[0]
    x2 = bb[1]
    y1 = bb[2]
    y2 = bb[3]

    # cutout image 
    image_cutout = image[y1:y2, x1:x2, ...]

    # here do some operation on image_cutout on the gpu

    # paste it back
    image[y1:y2, x1:x2, ...] = image_cutout

Is there some way to vectorize this? I would like to, having all the image cutouts, put them back into original image. Thanks for any help!