SOSONAGI / SAM-MLX

one-click segment anything via gradio app

SAM-MLX

one-click segment anything via gradio app

You can attach image and segment anything from the SAM-VIT model.

How to

1. Clone repository from the mlx and mlx-examples

1-1 clone the mlx

git clone https://github.com/ml-explore/mlx.git

1-2 clone the mlx-examples

git clone https://github.com/ml-explore/mlx-examples.git

then go to segment_anything path

2. Install the requirements

pip install mlx==0.15.0
pip install mlx_lm
pip install gradio
pip install opencv-python

3. You have to download app.py into segment_anything path

4. Download the model or convert the mlx format from the sam-vit-base or other type model

4-1 Download link

git clone https://huggingface.co/sosoai/sam-vit-base-mlx

or refer to below mlx_lm links for your own model convert

https://github.com/ml-explore/mlx-examples/tree/main/segment_anything

This still needs lots of things for development in that output of segment is not correct as notebooks.

Really hope for contribution.

TO DO

1. Make the app in gradio (V)
2. share in github (V)
3. improve the performance for segment and box

Licenses Followed everythings from the SAM model (Apache 2.0) and MLX.