Skip to content

AzemaBaptiste/pokemon_classifier

Repository files navigation

                          

Bazema_pokemon

Project overview

This project aims to classify images as Pokémon. When the input image is a Pokémon, the algorithm will respond the Pokémon name. When the input image is a human face or anything else, the algorithm will respond which Pokémon you lookalike.

Inspired by CNN Project: Dog Breed Classifier

Technical environment

  • OpenCV
  • PyTorch
  • CNN
  • Transfer Learning
  • Image classification
  • Colab

Datasets

  • Labeled Faces in the Wild Home. This dataset will be used to evaluate a human face detector.
  • Pokemon Generation One. This dataset will be used to train and evaluate a Pokémon detector and a classifier of Pokémon species.
  • CIFAR-100. This dataset will be used to train the Pokémon detector, providing examples not representing Pokémon.

Training workflow

train.svg

Prediction workflow

predict.svg

Usage

Docker (recommended)

# Assuming you have an image named "pika.jpg" in the current directory
docker run -v $PWD:/app bameza/bazema_pokemon:latest --image_path pika.jpg

Local development

Requirements

Python >= 3.6

virtualenv -p python3.8 venv && source venv/bin/activate
make install
make linter

bazema_pokemon --image_path pika.jpg

Resources

Todo

  • make docker image smaller

About

No description or website provided.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors