README.md

metadata

tags:
  - image-classification
  - imagenette
  - resnet50-pytorch
  - pytorch
datasets:
  - imagenette
metrics:
  - accuracy

ImageNette resnet50-pytorch

Model Description

resnet50-pytorch trained on 10-class ImageNet subset (ImageNette) with advanced augmentation techniques.

Model Architecture

• Architecture: resnet50-pytorch
• Dataset: ImageNette
• Classes: 10

Training Configuration

• Batch Size: 128
• Optimizer: sgd (momentum=0.9, weight_decay=1e-3)
• Scheduler: onecycle
• Augmentation: HorizontalFlip, ShiftScaleRotate, Cutout, ColorJitter
• MixUp: Alpha=0.2
• Label Smoothing: 0.1
• Mixed Precision: True
• Gradient Clipping: 1.0

Performance

• Best Test Accuracy: 24.28%
• Total Epochs Trained: 2
• Final Train Accuracy: 16.16%
• Final Test Accuracy: 24.28%

Training History

• Best Epoch: 2
• Train Loss: 2.4702 → 2.3422
• Test Loss: 2.6904 → 2.1273

Usage

import torch
from huggingface_hub import hf_hub_download

# Download model
checkpoint_path = hf_hub_download(
    repo_id="pandurangpatil/imagenet10trial",
    filename="best_model.pth"
)

# Load checkpoint
checkpoint = torch.load(checkpoint_path, map_location='cpu', weights_only=False)

# Load model (you'll need to have the model definition)
# from models import get_model
# model = get_model('resnet50-pytorch', num_classes=10)
# model.load_state_dict(checkpoint['model_state_dict'])
# model.eval()

Training Details

• Dataset: ImageNette (9469 train, 3925 test)
• Classes: 10
• Image Size: 160×160 or 224×224
• Normalization: mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)

Files

• best_model.pth - Best performing model checkpoint
• training_curves.png - Training/test accuracy and loss curves
• lr_finder_plot.png - Learning rate finder results
• metrics.json - Complete training history
• config.json - Hyperparameter configuration

License

MIT

Citation

@misc{resnet50-pytorch-imagenette,
  title = {ImageNette resnet50-pytorch},
  year = {2025},
  publisher = {HuggingFace},
  url = {https://huggingface.co/pandurangpatil/imagenet10trial}
}