The ground truth labels are not attached to the published dataset.
For each rgba file, there is a visualized segmentation file, but the colors are not consistent between the segmentation images.
For example, in toybox-5 scene 151 the color of "chair" and "car" are wrong, it happens in each scene where the "table" label is missing.
Can you please publish the ground truth labels use are using? ["1=airplane", "2=car", "3=chair", "4=sofa", "5=table"]
Hi,
The ground truth labels are not attached to the published dataset. For each rgba file, there is a visualized segmentation file, but the colors are not consistent between the segmentation images. For example, in toybox-5 scene 151 the color of "chair" and "car" are wrong, it happens in each scene where the "table" label is missing. Can you please publish the ground truth labels use are using? ["1=airplane", "2=car", "3=chair", "4=sofa", "5=table"]