Fetching the paper…

MCUNetV2: Memory-Efficient Patch-based Inference for Tiny Deep Learning · Around