Pokémon Image Classifier (Machine Learning)
A machine learning project for classifying Pokémon species from images using a supervised image classification model.
Overview
This project trains a deep learning model to recognize Pokémon from images. The model was trained using the dataset and training pipeline from the Pokémon Classifier project:
The goal of this project is to explore large-scale image classification and evaluate model performance on unseen Pokémon images.
Dataset
The dataset consists of 36,196 labeled images split into three independent sets:
| Split | Images |
|---|
| Training | 25,309 |
| Validation (Eval) | 5,439 |
| Test | 5,448 |
| Total | 36,196 |
The validation and test sets are completely separate from the training data to provide an unbiased measure of model performance.
Training Results
The current model achieved approximately:
- Training Accuracy: ~60%
- Validation Accuracy: ~80%
The higher validation accuracy indicates that the model generalizes well on the evaluation dataset. Future work will focus on improving training convergence and overall accuracy through additional hyperparameter tuning, data augmentation, and architecture improvements.
Machine Learning Pipeline
- Dataset preprocessing
- Image augmentation
- Supervised training
- Validation after each training epoch
- Final evaluation on an unseen test set
Dataset Credits
This project is based on the Pokémon Classifier dataset:
Please refer to the original repository for dataset generation details and licensing information.
Possible Improvements
- Increase training accuracy
- Experiment with different model architectures
- Hyperparameter optimization
- Additional data augmentation
- Transfer learning from larger pretrained vision models
Acknowledgements
- Pokémon Classifier Dataset: https://github.com/immacdonald/pokemon-classifier
- Nintendo, Game Freak, and Creatures Inc. for the Pokémon intellectual property.
Disclaimer
This is an educational machine learning project intended for research and experimentation. Pokémon and all related names, artwork, and trademarks are the property of Nintendo, Game Freak, and Creatures Inc.
