All PYNQ releases ship with the popular OpenCV library pre-installed. The PYNQ computer vision overlays enable accelerating OpenCV components in Programmable Logic (PL). These overlays expose a subset Xilinx' xfOpenCV library (a part of Xilinx' reVISION solution) at the Python level, combined with the support for HDMI input/output (Pynq-Z1 and Pynq-Z2 only). Webcam, stream or file based input/output remains available through the pre-installed SW OpenCV (on all Pynq boards).
Currently this package is compatible with PYNQ image v2.4.
A number of prebuilt overlays are available for your PYNQ-enabled board. For example:
- Pynq-Z1/ Pynq-Z2:
- 2D filter & dilate: accelerated 3x3 2D filter and 3x3 dilate. See filter2d_and_dilate.ipynb.
- 2D filter & remap: accelerated 3x3 2D filter and remap. See filter2d_and_remap.ipynb.
- Ultra96:
- 2D filter & dilate % absdiff. See 2__Overlays-filter2d_and_dilate.ipynb.
- remap & stereoBM: See 3a_Overlays-stereo_block_matching.ipynb.
- optical flow: See 4__Overlays-opticalflow.ipynb.
To install the computer vision overlay example on your board, open a terminal and run:
$ sudo pip3 install --upgrade git+https://github.com/Xilinx/PYNQ-ComputerVision.gitAfter the setup, new Jupyter notebooks will be added under the pynqOpenCV folder, ready to try out, no additional steps are needed.
When after profiling, the critical OpenCV modules in an application are identified, the generation of a tailored overlay is easily done by specifying those modules as a list in a CMake project file. After that, building the tailored overlay only requires running cmake to generate the Makefiles followed by make to build the overlay.
For detailed instructions: Building xfOpenCV Overlays for Pynq: CMake based sds++ cross-compilation. Currently, this only supports building on the Ultra96 with 2018.3 Xilinx tools.
Source files : 3-Clause BSD License
SDSoC Platform : Multiple License File
Sysroot
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent