What is your plan? You want to do coding CNN afresh? If so in C/C++ for ARM or for DSPs? If not, are you planning to use open source CNN for linux on ARM?

We will start with the simplest cnn example, the mnist. We will analyze it from the hardware point of view ( FPGA). Then we will trart to implement each layer and to compare it with the expected results from the python

You want to implement it yourself? or you have some students who have to do it? What is the FPGA board you wanna use?

I want to build it myself. I saw that you did some publications regards this topic before. Do you like to cooperate?

I am very busy this week. Let me think about it on Saturday. We can chat at that time then.

Looks great

I don't know much about your project by there are some resources I would like to recommend.

1. I'm interested in building ANN in FPGAs in general. I made an attempt in the past, but it was quite simple... Much later I found this:

Omondi, Amos R., Rajapakse, Jagath C., FPGA Implementations of Neural Networks

2. There is a joint collaboration between Google's Tensor Flow and ARM, who offer the CMSIS-NN. Here is something by them:

utensor-mnist-demo, https://github.com/uTensor/utensor-mnist-demo