[WIP] V2 refactor

[continue-revolution]

A complete refactor of the repository. Make code more organizable and cleaner. This update list is tentative.

New Models

All high-quality (>100 star w/ good performance) open-vocabulary detection / segmentation models or SAM based models are worth supporting

We need to support the following models (incomplete list)

• [x] Matting-Anything
• [x] FastSAM
• [ ] GroundingDINO new API (don't forget to remove cuda code)
• [ ] MetaCLIP
• [ ] HIPIE
• [ ] Semantic-SAM
• [ ] OpenSeeD
• [ ] OV-SEG
• [ ] SegFormer
• [ ] SAM-Fast
• [ ] EfficientSAM
• [ ] AnyDoor
• [ ] powerpaint
• [ ] Osprey
• [ ] CartoonSegmentation

Side Features

We need to support some other side features

• [ ] FP8
• [ ] UI zoom-in
• [ ] Unify box+point and front-end bounding box add / del / edit
• [ ] Prompt for each bounding box
• [ ] Automatic inpainting after t2i / i2i & t2v / i2v
• [ ] Registration at ControlNet forge
• [ ] Accompany my AnimateDiff extension
• [ ] Refactor code organization and README
• [ ] Use new APIs from CN like AnimateDiff
• [ ] inpaint hands

[x-CK-x]

@continue-revolution does the UI-zoom in feature come with the ability to adjust the radius size of dot/s we can place?

Also is it at all possible to allow the image to be placed on a larger canvas? maybe with a send to segment tab button in case it has to be put on a different tab?

Just wondering. idk how easy it is to allow the extension to have an extra tab for something like that to allow for a larger canvas to set the points on.