Quantized .onnx computer vision models that are smaller and faster than FP32 single-threaded while maintaining high accuracy. Every model runs on both ONNX Runtime and OpenVINO from the same file.