Woodscape Fisheye Semantic Segmentation for Autonomous Driving | CVPR 2021 OmniCV Workshop Challenge Forum

Go back to competition Back to thread list Post in this thread

> Submission Semantic Segmentation masks

"Semantic Segmentation masks should be saved as PNG images with each pixel containing one of the following values to represent the corresponding class"
Are the pixel values supposed to be [0,0,0] , [1,1,1] and so on or according to the 'seg_annotation_info.json' in the google drive folder?

Posted by: namangup @ April 27, 2021, 4:05 a.m.

Dear namangup,

Please create PNGs same as the size of the corresponding image and fill each pixel with [0, 0, 0], [1, 1, 1], ... or [9, 9, 9]. Only the integer in the first channel is considered, while the values in the rest of the channels are ignored.
The seg_annotation_info.json provides information on how the RGB images of segmentation masks are coloured. Please do not use JSONs as submissions with JSONs are not considered.

We will upload the baseline results and make public the baseline submission.zip shortly. Thanks.

Posted by: omnicv_2021_challenge @ April 28, 2021, 11:38 a.m.

I made a submission but the status is still submitted. No further processing like is there

Posted by: hpysr_s @ May 6, 2021, 10:51 a.m.
Post in this thread