Analyzing Tradeoffs between Activation Functions and Optimizers in First Gen. Pokemon Classification
The program is designed to analyze pictures of Pokémon TCG 1st generation cards. It aims to analyze the tradeoff between combinations of activation functions and optimizers, emphasizing the importance of well-performing variants in a pure, CNN-based image classification task.
To see:
- the entire procedure go to Trade Off Analyzing with Plateau notebook
- more information about the used techniques go to the documentation
Previous trainings and trials before TradeOff Analyzing with Plateau, which include various versions with different hyperparameters and schedulers.
Python snippets taken from the notebook
Pictures used in the paper and in this README.