The 3rd Large-scale Video Object Segmentation - Track 3: Referring Video Object Segmentation Forum

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

> Many team got results of 0.070, is this the png format bug?

To save the prediction mask, I save binary numpy array with PIL with YouTube-VOS palette to a png file. When I upload to the server, the results is 0.070 on both J and F. And some other teams have the same results. Is this the format problem? Is there any problem in my process?

Posted by: brianzhang12 @ Nov. 25, 2021, 2:42 a.m.

Please check the sample files of "valid_submission_sample.zip" of the data file and align the format. It seems that there are no issues in the server as other teams are scoring.

Posted by: seonguk @ Dec. 7, 2021, 1:46 p.m.
Post in this thread