Question about interactive segmentation evaluation.

[CoderZhangYx]

Thanks for authors' nice work. Could you please provide the implementation details of Table 2?
To be more specific, I want to know :

1. which subset of OpenImages did you use? validation, test, or 800 samples?
2. how did you evaluate on ADE20k? Did you evaluate all the semantic masks?
3. You mentioned in your supp that the COCO metrics are evaluated on 600 items , but the number of items I downloaded from RITM contain 800 items.

Expecting your responce, thanks!