https://github.com/mikeesto/sam2-coreml-python
Run Segment Anything 2 (SAM 2) on macOS using Core ML models
https://github.com/mikeesto/sam2-coreml-python
coreml python sam2 segment-anything
Last synced: 2 months ago
JSON representation
Run Segment Anything 2 (SAM 2) on macOS using Core ML models
- Host: GitHub
- URL: https://github.com/mikeesto/sam2-coreml-python
- Owner: mikeesto
- Created: 2024-12-30T10:51:05.000Z (5 months ago)
- Default Branch: master
- Last Pushed: 2025-01-06T06:00:36.000Z (5 months ago)
- Last Synced: 2025-01-23T11:36:13.786Z (4 months ago)
- Topics: coreml, python, sam2, segment-anything
- Language: Python
- Homepage:
- Size: 12.7 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Segment Anything 2 - Core ML
Run Segment Anything 2 (SAM 2) on macOS using Core ML models.
## Installation
1. Clone this repository
2. Install dependencies:
```bash
pip install coremltools numpy pillow opencv-python
```
## Directory Structure
```
sam2-coreml-python/
├── models/
│ ├── SAM2_1SmallImageEncoderFLOAT16.mlpackage
│ ├── SAM2_1SmallPromptEncoderFLOAT16.mlpackage
│ └── SAM2_1SmallMaskDecoderFLOAT16.mlpackage
├── script.py
└── README.md
```
## Usage
1. Download the Core ML models and place them in the `models` directory
2. Place your input image in the project directory
3. Update the `script.py` file with the input image path
4. Run the script:
```bash
python script.py
```
The script will generate `output_mask.png` containing the segmentation mask and `output_segmented.png` containing the segmented image.
## Models
The script expects the SAM 2 Core ML models. These need to be downloaded separately and placed in your file system following the above directory structure.
You can find the models here on [Hugging Face](https://huggingface.co/collections/apple/core-ml-segment-anything-2-66e4571a7234dc2560c3db26). I used the `coreml-sam2.1-small` models on my MacBook Pro M3 and the inference time was around 4 seconds. The script will work with other models as well.
## Credit
I took regular inspiration when writing this script from the implementation of [sam2-studio](https://github.com/huggingface/sam2-studio), a SwiftUI app that uses the same Core ML models.