Deep NN on a microcontroller

I am interested in putting a NN (a small CNN) onto a microcontroller (say a cortex m7). My question is: is there an active open source effort in this? There are (experimental) PyTorch resources for quantizing filtering the net to use 8-bit weights, and then one can export it to ONNX. From there, it appears that the largest project was ai-techsystems deepC that translates the model to C code that can be compiled for a particular microcontroller. But it doesn’t …

cnn cortex deeplearning experimental export filtering onnx open source pytorch resources small