metadata

license: mit
dataset_info:
  features:
    - name: image
      dtype: image
    - name: label
      dtype:
        class_label:
          names:
            '0': alakazam
            '1': arcanine
            '2': beedrill
            '3': clefable
            '4': cubone
            '5': dodrio
            '6': dragonite
            '7': electrode
            '8': farfetch'd
            '9': gyarados
            '10': hitmonchan
            '11': hypno
            '12': jigglypuff
            '13': jolteon
            '14': kingler
            '15': machop
            '16': magnemite
            '17': mew
            '18': mewtwo
            '19': nidoking
            '20': parasect
            '21': pidgey
            '22': poliwrath
            '23': primeape
            '24': psyduck
            '25': raichu
            '26': rattata
            '27': scyther
            '28': seaking
            '29': spearow
            '30': tangela
            '31': vileplume
            '32': wartortle
            '33': wigglytuff
            '34': zapdos
  splits:
    - name: train
      num_bytes: 3045750.3061224497
      num_examples: 171
    - name: val
      num_bytes: 679136.3469387755
      num_examples: 37
    - name: test
      num_bytes: 677220.3469387755
      num_examples: 37
  download_size: 4413352
  dataset_size: 4402107.000000001
task_categories:
  - image-classification
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-00000-of-00001.parquet
      - split: val
        path: data/val-00000-of-00001.parquet
      - split: test
        path: data/test-00000-of-00001.parquet

Dataset Card for Dataset Name

Generation I pokemon images dataset (35 pokemon).

Dataset Details

This dataset was curated using pokemon images found on bulbapedia. It contains 35 randomly selected Generation I pokemon, with 7 images per pokemon.

Dataset Description

The dataset has train/test/val splits. The train split contains 4 images per pokemon, the test & val splits contain either 2 or 1 image per pokemon.

The following images were selected from Bulbapedia:

the main pokemon image, as shown on the bulbapedia page for that pokemon
the Pokemon Red & Blue artwork
the Pokemon Red & Green artwork
the Pokemon Smile artwork
the Pokemon Original Series anime artwork
the Pokemon X&Y and/or Ruby & Saphire anime artwork
the Pokemon Global Link artwork

Dataset Labels

['alakazam','arcanine','beedrill','clefable','cubone','dodrio','dragonite','electrode',"farfetch'd",'gyarados','hitmonchan','hypno','jigglypuff','jolteon','kingler','machop','magnemite','mew','mewtwo','nidoking','parasect','pidgey','poliwrath','primeape','psyduck','raichu','rattata','scyther','seaking','spearow','tangela','vileplume','wartortle','wigglytuff','zapdos']

Dataset Sources [optional]

Repository: Bulbapedia/Bulbagarden

Uses

Direct Use

Intended for use in few-shot learning AI models

Out-of-Scope Use

[More Information Needed]

Dataset Structure

Data contains a train, test and validation split

Dataset Creation

Curation Rationale

Other pokemon classification datasets were too large and diverse for few-shot learning use cases.

Datasets:

VictoriaDerks
/

pokemon-classification