TinyML AI Model Zoo for Embedded Hardware

Our project focuses on creating a repository of TinyML-based AI models specifically designed for embedded hardware devices. Traditional AI models, typically run on powerful GPUs, are often inefficient on commercial embedded MCUs and MPUs due to the extensive floating-point calculations required. Our ModelNova model zoo bridges this gap by miniaturizing popular open-source models such as Ultralytics YOLO, Meta Segment Anything Model, and Microsoft PHI-3. Through techniques like quantization, pruning, and custom NPU compilation, we adapt these models to run efficiently on small devices. This ready-to-use model repository empowers engineers to quickly deploy AI systems on their hardware, facilitating rapid prototyping and accelerating time-to-market for high-impact solutions.