Maple

Maybe you could refer to [this issue](https://github.com/hustvl/ViTMatte/issues/4).

If you mean generate from alpha mattes for training or generate from a segmentation map like MatAny, it is possible. However, if you mean automatic object detection, I'm afraid not....

@LukaGiorgadze In ViTMatte, we generate in this way: https://github.com/hustvl/ViTMatte/blob/c2605975063bacd49852b9709cc1e0ebe58e4972/data/dim_dataset.py#L509

@LukaGiorgadze Thanks 😄

For me, I may try to generate only one trimap for both of them instead of two separate ones. You can achieve this easily by our matte-anything.

Do you mean something like this?  

I see. Interesting perspective. However, it seems difficult to the matting models like ViTMatte. Since the training framework is different. From my own perspective, it is also difficult to say...

I think it may be because of the padding. Check [here](https://github.com/hustvl/ViTMatte/blob/main/modeling/meta_arch/vitmatte.py#L57).

We use the grid sample strategy to reduce the inference computation burden. Please take a look at the last section of our paper for detail. The code can be found...

Yes. Or you can replace only the [forward function in the Block](https://github.com/hustvl/Matte-Anything/blob/main/modeling/backbone/vit.py#L249). BTW, when you reproduce the results on Distinctions-646, the results will be influenced by the different trimaps you...