tillbeemelmanns / cityscapes-to-coco-conversion Goto Github PK

This project works. But I'm curious what the following message means. Could you tell me, please?
"Warning: invalid contours."

Hello, I am doing segmentation for road and would like to know how to convert the cityscapes dataset to coco for road segmentation. Thanks in advance.

I assume that this doesn't work for the gtBbox3d labels, right?
Is there any easy way for doing this?

Other than that, would it be good to add this here as a "feature"?

Best

while visulaizing, by running inspect_coco.py
File "C:\Users\mitta\AppData\Local\Programs\Python\Python311\Lib\random.py", line 369, in choice
if not seq:
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

Has somebody successfully converted cityscapes dataset filtering classes such as traffic lights, poles, etc? Indeed, I did the change of desired classes in category_instancesonly and also commented on the invalid contour warning. However, even if I did that, I still get .json output files with 0 annotations.

Would this also work for the coarse portion of the dataset ?

hi , thank you for the great work !
I'm beginner in computer vision
how can i get the pixels that belongs to a specific instance in the image ?
i wanted to mask out the instance from the original image :
https://tillbeemelmanns.github.io/assets/images/convert-cityscapes-to-coco/preview.png
to this : https://i.ibb.co/g4XQfv1/city-to-coco-1.jpg
all pixels that inside the mask
and this : https://i.ibb.co/s5GyH7v/city-to-coco-2.jpg
all the pixel that is inside the bounding box
is there anyway to extract these pixel directly from the instances id ?
Thanks in advance.