mornville / Pokedex_ML

CNN Model to identify Pokemon. Custom Dataset - 33, 358 Train and 7, 396 Test images(100*100)

Pokedex_ML

CNNs Project to identify Pokemon.

Custom Dataset - 33, 358 Train and 7, 396 Test images(100*100), 166 Classes of Pokemon.

About

• Gathered custom dataset using Bing Search Api in get_images.py and search_images_bing.py
• Cleaned the downloaded images which couldn't be opened using clean_data.py
• Used Transfer Learning on some Pre-Trained model like VGG16 and Mobile Net
• Total Classes of Pokemon = 166

VGG16 Model (Accuracy 51.23%)

MobileNet Model (Accuracy 63.237%)

View Code in Kaggle

https://www.kaggle.com/mornville/pokedex