The masks we should prepare should be amodal or modal?

[b-niu]

Dear author, thanks for such a wonderful work which may make great help to instance segmentation tasks.

I noticed that in the Readme,

Using your own dataset If using your own dataset to train or test, you need to make sure that it contains accurate modal annotations (masks are required >and categories are optional). Inaccurate modal mask annotations, e.g., COCO original annotaions that may have large margin >between masks of occluding objects, will result in unsatisfactory results.

However, the masks in "COCOA dataset proposed in [Semantic Amodal Segmentation]" are amodal.

I am confused that, should we prepare modal masks or amodal masks?

Thanks a lot.

[XiaohangZhan]

If you use your OWN dataset, you should prepare accurate modal masks. COCOA contains both modal and amodal masks, and the amodal masks are used in evaluation.

[b-niu]

Thank you @XiaohangZhan , modal masks should be much easier to be prepared compared to amodal masks.