/HLS_for_CNN

This repo has codes for hardware accelerator design for CNNs using high level synthesis from Altera.

Primary LanguageC

High Level Synthesis for Deep Learning.

This repo has codes for hardware accelerator design for Deep CNNs using high level synthesis from Altera.

Requirements:

  • Visual Studio 2013 (the project was developed using windows host). Porting to linux should be straight forward.
  • OpenCV 2 was used in the host program for image capture and processing.
  • Intel OpenCL SDK 14.0
  • The binary files provided in this repo are for DE5-Net board (Stratix V FPGA board from Altera). But if you want you can use the same codes to compile for any other board supporting OpenCL SDK.

Setup:

  • Clone this repo.
  • Download the test data (in binary format) from here and put it in the project directory.
  • Un-zip the downloaded alexNet_de5v2_data.zip file.
  • Folder structure should look like:
    • HLS_for_CNN
      • alexNet_de5v2
      • common
      • Model
      • Data
      • README.md