Artelico Models
a collection of 200+ computer vision models for on-device use
2 min readupdated
This repository contains over 200 computer vision models (object detection, classification, segmentation, face recognition, and depth estimation) converted into Core ML (for iPhone, iPad, and Mac) and ONNX (for browsers) formats. All computations run locally right on your hardware without sending images to remote servers. The collection is based on architectures from the Hailo Model Zoo and optimized for local execution using hardware acceleration (Neural Engine, WebGPU). The repository is useful for mobile and web developers who need to quickly integrate ready-made AI for visual data processing.
Key features
- Local execution right on the device (Core ML and ONNX Runtime Web) without sending pictures to a server.
- Over 200 ready-to-use models: from classic YOLO and ResNet to modern ViT and CLIP transformers.
- Hardware acceleration on Apple Silicon processors via the Neural Engine for Core ML.
- Browser support via WebGPU and WebAssembly using ONNX Runtime Web.
What you can do
- Integrate a YOLO object detector or face recognition models into an iOS or macOS mobile app using Swift.
- Add image processing, object segmentation, or depth estimation to a web application using ONNX models in the browser.
- Run feature extraction or text and image analysis using CLIP / SigLIP encoders on the user's device.
Example
let url = modelsFolder.appendingPathComponent("yolov8s.mlpackage")
let compiled = try await MLModel.compileModel(at: url)
let model = try MLModel(contentsOf: compiled, configuration: MLModelConfiguration())
Read more
DetailsFormats: Core ML packages (.mlpackage) and ONNX graphs (.onnx) with output mapping files.
Hardware support: Apple Neural Engine (Core ML), WebGPU, and WebAssembly (web version).
Notes: The first launch of a large model on a Mac can take up to several minutes due to Neural Engine compilation.
Licenses: Vary for each model (MIT, Apache-2.0, GPL-3.0, Non-commercial research only).










Comments