vit-food-classifier-3class

ViT-Chef — Fine-tuned Vision Transformer for Food Classification

A fine-tuned Vision Transformer (ViT) model trained to classify pizza, steak, and sushi images. Achieves 96% accuracy on the test set, demonstrating the strong performance of transfer learning for visual food recognition.

---

Model Details

• Base model: google/vit-base-patch16-224-in21k
• Input size: 224 × 224
• Classes: ["pizza", "steak", "sushi"]
• Accuracy: 96% (Test set)
• Training: 5-fold cross-validation with AdamW optimizer and early stopping
• Dataset: Custom curated set (225 train, 75 test)

---

Usage

from transformers import ViTImageProcessor, ViTForImageClassification
from PIL import Image
import torch

model = ViTForImageClassification.from_pretrained("archit/vit-chef")
processor = ViTImageProcessor.from_pretrained("archit/vit-chef")

image = Image.open("example.jpg")
inputs = processor(images=image, return_tensors="pt")

with torch.no_grad():
    logits = model(**inputs).logits
pred = logits.argmax(-1).item()

print(model.config.id2label[pred])

---

Results

---