Model Card for augmented-donut-pose

Model Details

• Model Architecture: resnet18
• Pre-trained: Yes
• Number of Classes: 2

Dataset Details

• Dataset Name: lay-or-stand-donuts-dataset
• Training Samples: 246
• Validation Samples: 42
• Test Samples: 72

Training Metrics

• Final Training Loss: 0.1101
• Validation Accuracy: 100.00%

Data Augmentations

This model was trained with advanced data augmentations including RandomResizedCrop, RandomHorizontalFlip, RandomRotation, ColorJitter, and RandomErasing to improve generalization.

Augmented Sample Images

This repository also contains a selection of augmented sample images generated during the data preparation phase, which can be found in the augmented_samples/ directory.

Usage

This model is a classification model trained to distinguish between two classes related to donuts (e.g., 'lay' or 'stand'). It was fine-tuned on the lay-or-stand-donuts-dataset dataset with enhanced data augmentations.