Rle_decode produces flipped mask image

Are we missing transpose here?

plt.imshow(img)
mask_img = vision.rle_decode(mask, shape).astype(np.uint)
plt.imshow(mask_img[:, :, 0])

Unknown

I guess there may be two conventions for rle mask depending on if you put rows or columns first.

Yep, makes sense.

Hi Polakowo,

I am facing the same issue, getting a mask that fits the transpose of the image, How did you fix this…?

Regards,
Rony

Just transpose the output (image.T is the command)