More Accessible Open Source?

[zcai0612]

Thanks a lot for your brilliant work!

The two-stage pipeline of PICTURE has various inputs:

1. The input image.
2. The predicted human parsing (attribute).
3. The predicted human parsing (attribute) without cloths (with mask).
4. The predicted (extracted) DensePose.
5. The input image (or text prompt) of the target cloth shape.
6. The added mask applied to the input image as 3.
7. The target cloths.

So many inputs should be got when inferencing PICTURE, it is hard for someone wanting to apply it to their own data.

To make this work more accessible to others, do you have plans to release the code to get all the inputs?