Picture in rgb mode convert to L mode, Problem with soft picture edges

1

Hi there,
I’am working on an Object Detection with my own Data. Now I have the Problem than my pictures of the masks are in RGB mode but I need to convert them to L Mode. I have done thins with .convert(‘L’). But now I have a Problem with my edges. If you look at this picture

Maske_konvertiert
Maske_konvertiert1313×731 22.8 KB

the edges are not yellow. Do you have an idea how to convert it correctly? Before converting the picture was just black and yellow.
Thanks for help.

2

This issue doesn’t seem to be directly related to PyTorch, so you might get faster and better answers on e.g. StackOverflow.

That being said, I guess that that original image might also contain these artifacts.
If the input image was just a binary image encoded as an RGB image, you could apply some smoothing and binarization to try to get rid of these noisy edges.

3

Thanks for answer. I think I found the problem:
I set the colors of the picture in python, there all Pixels have the correct color. But if I save the picture as a jpg to download, the colors get wrong. Do you know how to fix this issue?

4

JPEG uses an internal encoding, which might create these artifacts. I don’t know which software or platform you are using, but you might be able to increase the quality level.
If that’s not possible, try to download these pictures in another format.