Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/olokobayusuf/segment-anything-unity
Segment Anything in Unity Engine, then run everywhere with NatML.
https://github.com/olokobayusuf/segment-anything-unity
Last synced: 26 days ago
JSON representation
Segment Anything in Unity Engine, then run everywhere with NatML.
- Host: GitHub
- URL: https://github.com/olokobayusuf/segment-anything-unity
- Owner: olokobayusuf
- License: apache-2.0
- Created: 2023-04-11T13:14:50.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2023-04-12T03:23:54.000Z (over 1 year ago)
- Last Synced: 2024-10-15T21:04:45.937Z (about 1 month ago)
- Language: Jupyter Notebook
- Homepage: https://ai.facebook.com/blog/segment-anything-foundation-model-image-segmentation
- Size: 29.3 KB
- Stars: 4
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Segment Anything in Unity
This is a realtime implementation of Meta AI's [Segment Anything Model](https://ai.facebook.com/blog/segment-anything-foundation-model-image-segmentation) in Unity Engine using NatML.## How it Works
The SAM projects consists of two models:
1. A large image embedding model
2. A lightweight mask decoder networkWe setup the embedding model using a NatML endpoint (see [embedding.ipynb](Endpoints/embedding.ipynb)). We export the lightweight mask decoder network to ONNX, then use NatML to convert it to CoreML for iOS; ONNX for Windows and the web; and TensorFlow Lite for Android.
## Requirements
- NatML 1.1.4+## Supported Platforms
- Android 24+
- iOS 14+
- macOS 10.15+
- Windows 10+ (64-bit only)
- Web:
- Chrome 91+
- Firefox 90+
- Safari 16.4+