Draft: Self-driving SuperTuxKart with Pytorch

  • Designed deep networks for a racing simulator, SuperTuxKart with Pytorch
  • Trained linear model and multi-layer perceptron model to classify images from SuperTuxKart
  • Trained a convolutional network to classify images from SuperTuxKart
  • Built classification network fully convolutional and solved a semantic labeling task (labeling every pixel in the image)
  • Implemented an object detector
  • Trained a CNN to do vision-based self-driving in SuperTuxKart
  • Programmed a SuperTuxKart ice-hockey player (AI that plays ice-hockey)