README.md

---
license: apache-2.0
base_model: dhkim2810/MobileSAM
tags:
  - mask-generation
  - vision.cpp
pipeline_tag: image-segmentation
---
# GGUF models for MobileSAM

MobileSAM is a model for image segmentation. It generates object masks from
point or box prompts. The weights in this repository are converted for
lightweight inference on consumer hardware with
[vision.cpp](https://github.com/Acly/vision.cpp).

* Original repository: [ChaoningZhang/MobileSAM (Github)](https://github.com/ChaoningZhang/MobileSAM)
* Original weights: [dhkim2810/MobileSAM (HuggingFace)](https://huggingface.co/dhkim2810/MobileSAM)

## Run

Example inference with [vision.cpp](https://github.com/Acly/vision.cpp):

```sh
vision-cli sam -m MobileSAM-F16.gguf -i input.png -p 256 480 -o mask.png --composite output.png
```

## Models

| Model                                    | Description                                                      |
| ---------------------------------------- | ---------------------------------------------------------------- |
| [MobileSAM-F16.gguf](MobileSAM-F16.gguf) | Encoder + decoder, fused batch norm, NHWC memory layout, float16 |