Fast object recognition of road objects with a deep neural neural network. You can preview results in real time in GUI, generate video with results or save all frames as images. Coloring scheme should be compatible with CityScapes dataset.
You just need to copy your own pretrained model and its description to ./models
To run with GUI you need to install OpenCV with GUI support.
If GUI is not necessary, pip install opencv-python may be sufficient.
For instructions on how to setup hipCafee, please see https://github.com/ROCmSoftwarePlatform/hipCaffe/blob/hip/README.ROCm.md
Set paths and parameters in main function, then
python run_demo.py