ZCU104 1D signal deep learning

Hello, I recently got interested in FPGAs and bought a zcu104.
I also installed PYNQ v2.6 and confirmed that it works normally.

I am planning to receive the ECG signal from the Galaxy Watch using socket communication, and I have confirmed that socket communication between the PC and the board is possible.

The ECG signal is a 1D signal, but the example uses deep learning such as CNN for 2D signals.

Are there any other examples using 1D signals?

I am using PYNQ for the first time. Is it correct to make the logic myself and run deep learning using Python on top of it?


See tutorial


Thank you for answer.

The tutorial you provided is 2D classification using the MNIST data set. I want a 1D signal.


First welcome to the community, but your modus is highly Inappropriate.
This is not a forum for seeking works nor content extraction.
If you are consider such contents GitHub or Google is a better place for you.
Help post is to report PYNQ BUGs, PYNQ syntax support etc. Not ZYNQ projects seeking.

From what you had described, I highly suggest you do more study on convolution and understand feature map and differences between 1D or 2D or even N-D neural network.

Kindly suggestion, if the action is pure FC behavior SIMD is always the fastest rather than using PL to compute.
As ARM can run over 600MHZ / DATA SET on SIMD.



I checked the link you provided.

Can I run it referring to PYNQ-2.7-MNIST-CNN/jupyter/sanity_test_hls.ipynb in the attached file?

Additionally, is there a tutorial on the most basic deep learning using PYNQ?


I cannot see how MNIST is not simply enough.
MNIST numeric conv neural network is almost >20years old.
Network is simple structure is also covered most knowledge about neural network behavior.
EGG would require a lot of noise filtering and other signal processing knowledge which CNN itself is also a kind of signal processing methodology.

I am sure old man like me also able to learn ChatGPT, you must able to ask ChatGPT on these contents as well.



A nice sharing on how training works on NN.
There are so many ways but all concepts are similar:

Adam, SGD etc.