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

Loss: 0.2039
Accuracy: 0.9269

Model description

Transfer-Learned ViT Model

This model was trained using transfer learning based on the ViT model google/vit-base-patch16-224.

🔧 Training Setup

Dataset: Oxford-IIIT Pets
Epochs: 7
Batch Size: 8
Learning Rate: 2e-4

Performance on Test Set

Accuracy: 0.9269
Precision: 0.9273
Recall: 0.9269

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
num_epochs: 7

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.3448	1.0	739	0.2690	0.9418
0.2359	2.0	1478	0.2013	0.9378
0.1621	3.0	2217	0.1807	0.9391
0.1436	4.0	2956	0.1738	0.9378
0.1106	5.0	3695	0.1679	0.9445
0.1319	6.0	4434	0.1616	0.9405
0.1413	7.0	5173	0.1609	0.9391

Framework versions

Transformers 4.50.0
Pytorch 2.6.0+cu124
Datasets 3.4.1
Tokenizers 0.21.1

Comparison with the zero-shot classification model

Zero-Shot Evaluation on the Oxford-IIIT Pet Dataset

As part of the evaluation, I compared this transfer-learned model to a zero-shot classification model. The zero-shot model used is:

Model: openai/clip-vit-large-patch14
Task: zero-shot-image-classification

Zero-Shot Model Results

On the Oxford-IIIT Pet Dataset, the zero-shot model achieved the following performance:

Accuracy: 0.8800
Precision: 0.8768
Recall: 0.8800

These results were obtained using the test set and evaluated with sklearn.metrics.

chrisis2/vit-base-oxford-iiit-pets

vit-base-oxford-iiit-pets