Dive into the advanced super-resolution technology of Dynamic Zoom, a real-time video enhancement system designed for digital media.
- Website: https://dynamic-zoom.github.io/home/
- Repository: https://github.com/Dynamic-Zoom/Dynamic-Zoom
- Ensure Conda (Miniconda/Anaconda) is installed on your system. (Official Conda website for installation guide)
- Run
conda env create -f environment.yml
to install necessary packages. - Activate your new Conda environment using
conda activate dynamic-zoom
. (Check for appropriate PyTorch version installation to enable GPU acceleration) - Run
python run.py --help
to check Command line arguments - Example run:
python run.py --input "data/input.mp4" --output "output/output.mp4"
Argument | Description | Supported values |
---|---|---|
-h, --help | Show help message and exit | |
-r, --roi | Define size of your ROI window | 2 (space separated) numbers: H W |
--input | Relative File path of the input file | |
--output | Relative File path of the output file | |
-m, --model_name | Select upscaling model (default = bicubic++) | "bicubic++" "srgan2x" "srgan4x" |
--fps | Fps of output file (default = 24) | |
-v, --verbosity | Select output verbosity (default = 0) | 0 = only pipeline stage logs(default) 1 = per frame logs 2 = detailed logs |
- Bicubic++ - by Aselsan Research group: https://github.com/aselsan-research-imaging-team/bicubic-plusplus
- Swift-SRGAN - by Koushik Sivarama Krishnan, Karthik Sivarama Krishnan: https://github.com/Koushik0901/Swift-SRGAN