A real time Histogram of Oriented Gradients Implamentaion on FPGA
Board used: Zedboard
Vivado Version: 2016.04
Brief description of project: In this project a real time implementation of the Histogram of Oriented Gradients pedestrian detection algorithm is presented. We have designed hardware accelerators via Vivado HLS in order to reduce the computation time of the HOG exctraction and classification. Furthermore, we have implemented an embedded application running on Petalinux with an Ubuntu image, that captures frames from the webcam connected to the Zedboard, controls and monitors the accelerators and renders the detections image to a VGA monitor.
Description of archive (explain directory structure, documents and source files):
Hog_HLS: Vivado_HLS project
Hog_System_Zynq: Vivado project
hw: bitstream, hardware description file and kernel configuration files (in case someone wants to build a fresh petalinux project from the beginning)
ip: Accelerator and 3rd party Display Controller IPs
sd_image: download the BOOT and rootfs files from: https://www.dropbox.com/sh/ehcngof3dpio38b/AAArrgH8EgVosfa9s2bgct3na?dl=0
src: src/c: Vivado HLS code and testbench. src/test_image: An image for testing purposes.
sw: The executable and the code of the embedded software application.
Instructions to build and test project
Accelerator test: In order to test and simulate the accelerators functionality we provide the HLS project.
System test: First of all a usb webcam with UVC driver is required and it is connected to the USB OTG port of the Zedboard. In addition, a VGA monitor should be connected to the VGA port of the Zedboard.
A build of the Petalinux is not required since we provide all the necessary files to build and run our application. To test it:
Step 1: Download the files from: https://www.dropbox.com/sh/ehcngof3dpio38b/AAArrgH8EgVosfa9s2bgct3na?dl=0.
Step 2: A SD card 4Gb or greater is required. It should be partitioned into two partitions so that the first one contains the contents of the BOOT folder and the second one the extracted contents of the rootfs folder. The SD card is now ready to be inserted into the Zedboard.
Step 3: You can either compile the application on the Zedboard using the provided makefile or run the executable application provided in sw/ directory.
