ailia-models: a verified library of 400+ pre-trained AI models optimized for high-speed cross-platform inference

What it solves

This repository provides a comprehensive collection of pre-trained, state-of-the-art AI models optimized for use with the ailia SDK. It solves the difficulty of finding, exporting, and deploying high-performance AI models across diverse hardware platforms (Windows, Mac, Linux, iOS, Android, Jetson, and Raspberry Pi).

How it works

Models from various frameworks (primarily PyTorch, but also Keras, TensorFlow, and OpenVINO) are exported and optimized for the ailia SDK. The SDK utilizes GPU acceleration via Vulkan and Metal to ensure high-speed inference across different operating systems and hardware, including non-OS/RTOS environments.

Who it’s for

Developers building AI-powered applications who need a verified library of models for fast, cross-platform deployment without having to manually optimize models for every target device.

Highlights

• Extensive Model Library: Over 400 verified models covering action recognition, anomaly detection, audio language models, and detailed audio processing.
• Broad Hardware Support: Optimized for GPU acceleration on a wide range of devices from mobile phones to single-board computers like Raspberry Pi.
• ** uma Multi-language Bindings**: Support for C++, Python, Unity (C#), Kotlin, Rust, and Flutter.
• Diverse AI Capabilities: Includes specialized models for speech-to-text, text-to-speech, speaker diarization, and industrial anomaly detection.