resnet50-finetuned-cifar10-classification

CIFAR-10 Image Classification with ResNet-50

This model fine-tunes microsoft/resnet-50 on the CIFAR-10 dataset using PyTorch. It is designed for robust image classification across 10 everyday object categories.

Model Details

• Base Model: microsoft/resnet-50
• Library Used: PyTorch
• Pipeline Tag: image-classification
• License: MIT

Training & Evaluation

• Dataset: CIFAR-10 (60,000 images, 10 classes)
• Preprocessing: Images resized to 224×224, normalized with ImageNet statistics
• Augmentation: Random horizontal flip, crop, color jitter for training set
• Validation/Test: No augmentation, only preprocessing
• Optimizer: Adam (lr=0.0005, weight_decay=5e-4)
• Loss: CrossEntropyLoss with label smoothing (0.1)
• Scheduler: ReduceLROnPlateau (factor=0.1, patience=3)
• Regularization: Dropout (p=0.5) before final layer

Metrics

Metrics are computed on the held-out CIFAR-10 test set.

Classes

• airplane
• automobile
• bird
• cat
• deer
• dog
• frog
• horse
• ship
• truck

Citation

If you use this model, please cite the base model and this repository.

MIT License applies.