How do I implement my own trained neural network medical image classification) on the PYNQ-Z2?

Support
1

How do I implement my own trained neural network medical image classification) on the PYNQ-Z2?
is there any repo. or example explain in detail I am read BNN-pynq and FINN but until know I don’t find any example with medical images classification using pynq ? can anyone helps me please

2

Have you tried exploring Vitis AI and DPU IP in xilinx?

DPU for Convolutional Neural Network (xilinx.com)

You won’t find specific application, medical data on pynq, for that you have to look in two parts, medical data ML done by people around the world, and AI resources that exist for the pynq board.

3

Hi, while you have some resources on internet, there isnt much information about how to implement your own customized network easily. I had the same issue a time ago that most of the “tutorials” where very specific and if you wanted to do something different a lot of errors appeared.

You are trying to implement image classifications, do you have your network already trained? Are you using tensorflow, pytorch?

To convert a trained network to HLS and then to vivado you can try https://github.com/fastmachinelearning/hls4ml. It works on a pynq-z2 with a little work and it has a nice tutorial where you can learn how to use it.

The steps are:

  1. Train your network
  2. Convertit to HLS with HLS4ML and create your IP block
  3. Generate your vivado project
  4. Generate a bitfile and use it on the PYNQ.
1 Like
4

thanks alot sir