This repo uses Tensorflow DNNClassifier to determine a pokemon type based on its image. Was inspired by the original blog article from Juan De Dios Santos at https://juandes.com/pokemon-colors-and-deep-learning-95fb715be46
Total Sample size of images: 714 Training Size: 650 Test Size: 64
DNNClassifier uses 300 Hidden Units and trains on a Batch Size of 10 over 20000 Steps. Accuracy of the model on the Test Sample only at 0.15
PokemonCNN builds a CNN with 2 convolutional layers and 2 Pooling Layers.