PINTO_model_zoo

Why we included this project

Anyone who has hunted for a vision model already converted to their target runtime knows the drill: a TensorFlow Lite build sized for an EdgeTPU, an OpenVINO IR for an Intel stick, a CoreML package for iOS. This collection gathers inter-converted versions of the same networks, from EfficientNet and MobileNet to pose and segmentation models, across TensorFlow, PyTorch, ONNX, OpenVINO, TensorFlow.js, TF-TRT, CoreML, and quantized TensorFlow Lite variants. The ready-made INT8 and EdgeTPU builds matter most for edge hardware, since they spare you the conversion and calibration work. Each model folder carries its own license note and the conversion scripts, so you can re-derive a format instead of trusting the artifact blindly. Sample code is thin in places, but that gap is openly acknowledged and open to pull requests.