This is LeNet5 Accleration Code
Accelerated implementation using Zynq (Zc706) and SDSoC
Designed by Constant Park ([email protected])
This is original code of LeNet-5 Accleration with HLS.

You can read the detailed information at this paper.
- S.Park,＂Implementation of a CNN accelerator on an Embedded SoC Platform using SDSoC", ICDSP, 2018.
- C.Lee and S.Park,"Implementation of an FPGA-based CNN Accelerator using SDSoC", Annual Conference of the Korean Institute of Communication Sciences, 2017.

Copyright by Constant Park and Embedded SoC Lab (ESoC Lab). All implementation can be modified with any permission.
It does not matter if you can make your project. 

I give some Question and Answer.
Q: How can you train this model ?
A: I trained the LeNet-5 model using Tiny-DNN. Extract the weight from Tiny-DNN and Run this implementation on Zynq.
But weights in this project is not trained model. So You should train LeNet-5 model and extract weights from DNN Framkeworks.

Q: In your paper, the upper loop is applied by Pipeline and the lower loop is applied Unroll. But it is impossible in SDSoC. How can you do that ?
A: In SDSoC tool, Pipeline includes not only the pipeline technique, but also loop unrolling. If you apply a pipeline to the upper part and make it unroll well internally, both the pipeline and unroll apply.

Q: How can you measure the power consumption ?
A: Using power like this one (https://www.youtube.com/watch?v=afhtMWh5P0I)

Feature works
I will upload MobileNet-HLS Code and SRCNN-HLS Code.