Automatic grading and generation of moonboard problems.
This code implements the convolutional neural network for climbing route grade classification on the Moonboard found here. The code for this is found in grade_net.py. grade_net_comp.py implements another route climbing classification CNN with significantly more parameters. Both algorithms were found to perform similarly.
Additionally, an auxillary classifier generative adversarial network for generation of moonboard problems is implemented in moon_gan.ipynb. This allows us to assign a grade to each generated problem produced by the generator. An example of a generated problem with the proposed grade of 6C+ is shown below.
Generated problems can be viewed through compose_problems.ipynb after running the AC-GAN to generate a model.
$ git clone https://github.com/adamreidsmith/moon_ai
$ cd moon_ai/
$ sudo pip3 install -r requirements.txt
- Python 3
- Jupyter Notebook
- PyTorch
- OpenCV
- NumPy
- Matplotlib
- Seaborn