- Sample Code
- Task
- FastViT
- Image Classification
- Depth Anything V2 (small)
- Monocular Depth Estimation
- DETR (ResNet 50)
- Semantic Segmentation
Specs
- Sample CodeTask
- FastViTImage Classification
- Depth Anything V2 (small)Monocular Depth Estimation
- DETR (ResNet 50)Semantic Segmentation
README
CoreML Examples
This repository contains a collection of CoreML demo apps, with optimized models for the Apple Neural Engine™️. It also hosts tutorials and other resources you can use in your own projects.
Demo Apps
||||
| ------------- | ------------- | ------------- |
| <video src="https://github.com/huggingface/coreml-examples/assets/45471420/04f13819-bdae-4843-9631-940bd9b21b4e"> | <video src="https://github.com/huggingface/coreml-examples/assets/45471420/e760cf8b-0f11-4b95-9db6-db4c46d954d6">|<video src="https://github.com/huggingface/coreml-examples/assets/45471420/4b71b9e9-3d63-400c-868e-f0734aba0a6f">|
The models showcased include:
| Sample Code | Task | Core ML packages |
|---|---|---|
| FastViT | Image Classification | apple/coreml-FastViT-T8 apple/coreml-FastViT-MA36 |
| Depth Anything V2 (small) | Monocular Depth Estimation | apple/coreml-depth-anything-v2-small |
| DETR (ResNet 50) | Semantic Segmentation | apple/coreml-detr-semantic-segmentation |
We leverage coremltools for testing and implementing these optimisations. You can read more about it here.
Tutorials
- How to convert Depth Anything v2 for GPU and Neural Engine. Source code notebook <a target="_blank" href="https://colab.research.google.com/github/huggingface/coreml-examples/tutorials/depth-anything-coreml-guide.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/>
</a>