This is a vision design for 2021 Xilinx summer school, to implement and accelerate haze removal algorithm based on PYNQ platform. It includes the details from software algorithm to hardware acceleration.
You can stream the haze removal result of your laptop display to another monitor:
And also the depth map estimated by thickness of haze:
- Board: PYNQ-Z2
- PYNQ Image: PYNQ v2.6
- Install composable overlay
Browse the ipython notebooks in Jupyter Lab after clone.
Then copy the folder containing composable overlay bitstream into this repo.
- 0: some bugs (ignore it)
- 1: introduce and improve the DCP haze removal algorithm
- 2: process video stream with base overlay, using previous algorithm and interactive feature of ipthon
- 3: replace base overlay with composable overlay
- 4: accelerate haze removal with IP core generated by HLS