metadata

license: mit
tags:
  - image-classification
  - computer-vision
  - animals10
  - resnet
  - pytorch-lightning
datasets:
  - custom
library_name: pytorch-lightning

🐾 Animal Breed Classifier (ResNet18 - Animals10)

This model is a fine-tuned ResNet18 trained on the Animals-10 dataset using PyTorch Lightning. It classifies images into one of 10 animal categories.

🧠 Model Details

Architecture: ResNet18
Framework: PyTorch Lightning
Training dataset: Animals-10
Input size: 224x224 RGB
Classes: ['butterfly', 'cat', 'chicken', 'cow', 'dog', 'elephant', 'horse', 'sheep', 'spider', 'squirrel']

📊 Performance

Metric	Value
Accuracy	95%
Macro F1	94%
Weighted F1	95%

For a detailed comparison with zero-shot CLIP, see the GitHub repo.

🚀 Usage

from models.lightning_model import Classifier

model = Classifier.load_from_checkpoint(
    'https://huggingface.co/MichaelMM2000/animals10-resnet/resolve/main/resnet_animals10.ckpt',
    num_classes=10
)

🧪 Dataset

The Animals-10 dataset consists of over 28,000 images across 10 animal categories.

📎 License

This model is released under the MIT License.