Computer Vision Demo Framework

Dependencies

To build the core application, the following libraries are required:

• OpenCV version 3.1 or later (must build from source on Ubuntu 16.04)
• OpenCV extra modules 3.1 or later (must include during OpenCV build process)
• Boost version 1.58 or later (libboost-all-dev on Ubuntu 16.04)
• CMake version 3.2 or later

For video output support, GStreamer and associated plugins must also be installed before compiling CMake. On Ubuntu 16.04, the following packages should be installed:

• ffmpeg
• gstreamer1.0-plugins-base
• gstreamer1.0-plugins-good
• gstreamer1.0-plugins-ugly
• gstreamer1.0-libav
• libavcodec-dev
• libavcodec-ffmpeg56
• libavdevice-ffmpeg56
• libavfilter-ffmpeg5
• libavformat-dev
• libavformat-ffmpeg56
• libavresample-dev
• libavresample-ffmpeg56
• libavutil-dev
• libavutil-ffmpeg56

For FPGA-accelerated algoritms, the Altera SDK should be set up and accessible to CMake. The build scripts will automatically test the SDK installation before trying to use it, and display errors if it finds that any component is not set correctly.

Installation

To build the application, open a new shell in the source directory and run the following commands:

mkdir build && cd build
cmake .. && make -j 8

Once the build process finishes, you should have a compiled version of the core binary, pddemo, and any enabled libraries. If you're building the FPGA-based pedestrian detection demo and didn't get a copy of the bitstream from somewhere else, you should also build that by running make aoc. This does not run automatically because bitstream generation can take 5-7 hours on a fast machine.

Running

First, run ./pddemo --list-algos to display a list of available algorithms. If the one you want to show is not on this list, then something went wrong during another part of the build process, or required system libraries aren't properly installed.

Once you know the algorithm you want to run, use the -a [algorithm] flag to specify it. If no -a flag is given, the demo will default to using OpenCV's CPU-based pedestrian detection algorithm. To give a file or capture device as input, use ./pddemo [flags] file.avi or ./pddemo [flags] camera respectively. There is currently no supported syntax for using a video capture device other than the default.

Once the application is running, a window should pop up showing the current frame and the active algorithm's results. If no DISPLAY variable is present in the process's environment or if the -w flag is given, it will not atttempt to generate a window.