donkey-vis

Donkeycar neural network model visualization built on top of https://github.com/raghakot/keras-vis

Requirements

1. OpenCV
2. Conda
3. Trained neural network
4. Donkeycar records tub

Usage

1. Install opencv (OS Specific)
2. Install conda environment conda env create -f environment.yml
3. Activate conda env conda activate donkey-vis
4. For now: pip install git+https://github.com/raghakot/keras-vis.git --upgrade (as long as some bug has been fixed)
5. Check that the installed tensorflow version matches with the version model was trained on conda list | grep tensorflow (Defaults to 1.11.0)
6. If it doesn't match, update tensorflow version conda install tensorflow-gpu==1.11.0
7. Run visualization with python visualize --model ./model --tub ./records (python visualize -h for more parameters)
8. Kill with ctrl-c when enough frames exported