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

Train Loss: 0.2187
Validation Loss: 0.2654
Train Accuracy: 0.9420
Epoch: 5

Model description

This model leverages the Vision Transformer (ViT) architecture to classify retinal images for early detection of Diabetic Retinopathy (DR). The fine-tuned model improves accuracy and generalization on medical imaging datasets.

Intended uses & limitations

Intended Uses:

Medical diagnosis support for Diabetic Retinopathy
Research applications in ophthalmology and AI-based healthcare

Limitations:

Requires high-quality retinal images for accurate predictions
Not a substitute for professional medical advice; should be used as an assistive tool

Training and evaluation data

The model was trained on a curated dataset containing labeled retinal images. The dataset includes various severity levels of Diabetic Retinopathy, ensuring robustness in classification.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

optimizer: {'name': 'AdamWeightDecay', 'learning_rate': {'module': 'keras.optimizers.schedules', 'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 3e-05, 'decay_steps': 146985, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}, 'registered_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

Epoch	Train Loss	Validation Loss	Train Accuracy
1	0.4513	0.5234	0.8270
2	0.3124	0.4102	0.8930
3	0.2751	0.3856	0.9150
4	0.2376	0.3012	0.9320
5	0.2187	0.2654	0.9420

Framework versions

Transformers 4.46.2
TensorFlow 2.17.1
Datasets 3.1.0
Tokenizers 0.20.3