food-classifier

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the food101 dataset. It achieves the following results on the evaluation set:

• Train Loss: 0.2136
• Validation Loss: 0.2284
• Train Accuracy: 0.94
• Epoch: 4

Model description

This is an image classification model fine tuned from the Google Vision Transformer (ViT) to classify images of food.

Intended uses & limitations

For messing around!

Training and evaluation data

The training set contained 101 food classes, over a dataset of 101,000 images. The train/eval split was 80/20

Training hyperparameters

The following hyperparameters were used during training:

• optimizer: {'name': 'AdamWeightDecay', 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 3e-05, 'decay_steps': 20000, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'decay': 0.0, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False, 'weight_decay_rate': 0.01}
• training_precision: float32

Training results

Train Loss	Validation Loss	Train Accuracy	Epoch
0.3409	0.2903	0.932	0
0.2838	0.2897	0.917	1
0.2415	0.2869	0.914	2
0.2143	0.2630	0.924	3
0.2136	0.2284	0.94	4

Framework versions

• Transformers 4.30.2
• TensorFlow 2.12.0
• Datasets 2.13.1
• Tokenizers 0.13.3