This is an entry for this video by Siraj on Youtube.
The pokemon classifier aims to train a neural network to classify pokemon by their type 1 (i.e fire, water, grass, etc.) using this pokemon dataset on Kaggle.
- tensorflow (pip install tensorflow)
- numpy (pip install numpy)
Run the following in terminal
$ python main.py
or with all the variables defined
$ python main.py --verbose --trainingIterations 120 --learningRate 0.0005
The python script is able to parse the provided Pokemon dataset and train to an accuracy of around 75% after 120 iterations at a learningRate of 0.0005. After the training process, the user is then able to input their own Pokemon stats to see what the network thinks its type 1 is.
Credits go to Alberto Barradas (For the dataset), and Siraj (for the idea and starting code).