metadata

license: apache-2.0
tags:
  - image-classification
  - generated_from_trainer
metrics:
  - accuracy
widget:
  - src: >-
      https://datasets-server.huggingface.co/assets/keremberke/pokemon-classification/--/full/train/3/image/image.jpg
    example_title: Abra
  - src: >-
      https://datasets-server.huggingface.co/cached-assets/keremberke/pokemon-classification/--/full/train/383/image/image.jpg
    example_title: Blastoise
model-index:
  - name: pokemon_classifier
    results:
      - task:
          name: Image Classification
          type: image-classification
        dataset:
          name: keremberke/pokemon-classification
          type: pokemon-classification
          config: full
          split: validation
          args: full
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.08848920863309352
datasets:
  - keremberke/pokemon-classification
language:
  - en
pipeline_tag: image-classification

pokemon_classifier

This model is a fine-tuned version of google/vit-base-patch16-224 on the pokemon-classification and the full datasets. It achieves the following results on the evaluation set:

Loss: 8.0935
Accuracy: 0.0885

Model description

This model, referred to as "PokemonClassifier," is a fine-tuned version of google/vit-base-patch16-224 on Pokemon classification datasets. Its primary objective is to accurately identify the Pokemon in input images. While this general summary provides information about its performance in terms of loss and accuracy, its core function lies in precisely classifying Pokemon images.

Intended uses & limitations

This model is limited to the training data it was exposed to and can only identify the following Pokémon: Golbat, Machoke, Omastar, Diglett, Lapras, Kabuto, Persian, Weepinbell, Golem, Dodrio, Raichu, Zapdos, Raticate, Magnemite, Ivysaur, Growlithe, Tangela, Drowzee, Rapidash, Venonat, Pidgeot, Nidorino, Porygon, Lickitung, Rattata, Machop, Charmeleon, Slowbro, Parasect, Eevee, Starmie, Staryu, Psyduck, Dragonair, Magikarp, Vileplume, Marowak, Pidgeotto, Shellder, Mewtwo, Farfetchd, Kingler, Seel, Kakuna, Doduo, Electabuzz, Charmander, Rhyhorn, Tauros, Dugtrio, Poliwrath, Gengar, Exeggutor, Dewgong, Jigglypuff, Geodude, Kadabra, Nidorina, Sandshrew, Grimer, MrMime, Pidgey, Koffing, Ekans, Alolan Sandslash, Venusaur, Snorlax, Paras, Jynx, Chansey, Hitmonchan, Gastly, Kangaskhan, Oddish, Wigglytuff, Graveler, Arcanine, Clefairy, Articuno, Poliwag, Abra, Squirtle, Voltorb, Ponyta, Moltres, Nidoqueen, Magmar, Onix, Vulpix, Butterfree, Krabby, Arbok, Clefable, Goldeen, Magneton, Dratini, Caterpie, Jolteon, Nidoking, Alakazam, Dragonite, Fearow, Slowpoke, Weezing, Beedrill, Weedle, Cloyster, Vaporeon, Gyarados, Golduck, Machamp, Hitmonlee, Primeape, Cubone, Sandslash, Scyther, Haunter, Metapod, Tentacruel, Aerodactyl, Kabutops, Ninetales, Zubat, Rhydon, Mew, Pinsir, Ditto, Victreebel, Omanyte, Horsea, Pikachu, Blastoise, Venomoth, Charizard, Seadra, Muk, Spearow, Bulbasaur, Bellsprout, Electrode, Gloom, Poliwhirl, Flareon, Seaking, Hypno, Wartortle, Mankey, Tentacool, Exeggcute, and Meowth.

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.0872	0.82	500	7.2669	0.0640
0.1581	1.64	1000	7.6072	0.0712
0.0536	2.46	1500	7.8952	0.0842
0.0169	3.28	2000	8.0935	0.0885

Framework versions

Transformers 4.30.2
Pytorch 2.0.1+cu118
Datasets 2.14.5
Tokenizers 0.13.3