Deeplearning project done in collaboration with Morten Omholt-Jensen. The project was done for a master degree course, IMT4392 Deep learning for visual computing, at NTNU Gjøvik. The main idea of the project was to see how well a neural network could predict the outcome of a round of CSGO solely based on what equipment each team brought into the round.

The project includes three main parts:

  1. Datascraper and demo-info extractor to create dataset (Golang)
  2. OpenCV template matching to read equipment on teams from a CSGO match image (Python)
  3. Deep-learning neural network imlpementation with Keras and Tensorflow in Python to predict which team wins the round

Our final trained model was able to predict outcome of rounds with an accuracy of ~68%.

Repository can be reached at: https://github.com/mortenoj/deep-learning-project/

Click here if you want to see the full project report.

Image taken from stream at https://www.twitch.tv/starladder_cs_en.

Training session testing out different algorithms and hyperparameters to optimize for the best results.