Instance Segmentation masks are all 0 images

[utkutpcgl]

Thanks for your contribution.

Instance segmentation masks provided for your dataset (in the drive link) are all only 0-valued images (no mask available). Also, the links for instance, and normal segmentation labels should be swapped.

[ssnirgudkar]

I will check and will get back to you shortly. Thanks for reporting.

[ssnirgudkar]

I swapped the links of segmentation masks and instance masks. As for the instance segmentation masks, I checked the images. They are correct but the information is not sufficient. If you print the pixel values from these images, it will print the instance-ids of the classes. You will need mapping of these ids to RGB values. Let me see if I have that information. Meanwhile, can you share the details of your research and how you plan to use this dataset? Thanks.

[utkutpcgl]

I have the code to map ids to RGB values, thanks. I've checked the instance segmentation masks again, the mask of all images are all zero-valued pixels, rather than instance-ids (I am sure that there are images with multiple instances of a class). It would be nice if you could double-check the instance segmentation masks you have provided here: https://drive.google.com/file/d/16rPWhV8OEocyLLpT1-LSCALt757P_h0N/view?usp=sharing .

I will train a detection model on the dataset by converting segmentation masks to bounding boxes.

Thanks.