---
license: mit
task_categories:
- image-classification
language:
- en
tags:
- Pokemon
- Pokedex
- Image-Classification
size_categories:
- 1K<n<10K
---

# First Generation Pokemon Dataset

All images used to train and validate the classifier were collected by me using custom scripts from

* [Bulbapedia Archives](https://archives.bulbagarden.net/wiki/Main_Page)
* [Pokémon DB](https://pokemondb.net/)
* [Project Pokémon](https://projectpokemon.org/)
* [PTGigi Randomizer](https://ptgigi.com/apps/pokemon/randomizer/)

## Dataset Details

### Dataset Description

A total of 3,089 Pokémon images were collected from multiple websites, with the background removed from some of them. All images were then normalized to a resolution of 224x224 pixels.


### Dataset Structure

The dataset is divided into:

* Test: 840 images
* Train: 2249 images

#### Google Colab Notebook

Explore the model training and inference workflow in this interactive notebook:

* [Pokédex Colab Notebook](https://colab.research.google.com/drive/1IeGzndeOZ_9PnnWSnLjYgaS7QFn-4Voc?usp=sharing)

#### Pokédex Model

DenseNet121 neural network to create a Pokédex: 

* [First Gen Pokémon Model](https://huggingface.co/RogerKoala/gen1-pokemon-classifier)

#### Demo

Visit the live Space to try it out:

* [Hugging Face Space](https://huggingface.co/spaces/RogerKoala/Pokedex)