VamsiKari/food-vs-notfood-model

🍽 Dish Image Verification Model

This model verifies whether a given image matches a provided dish name.
It has been trained on a curated dataset of food images and labels, optimized for high precision in dish recognition tasks.

---

📌 Model Details

• Architecture: ResNet / CNN-based (PyTorch)
• Training Type: Fine-tuning on custom dataset
• Framework: PyTorch
• Task: Image classification (verification)
• Input: An image of a dish
• Output: Probability of the image matching the given dish name (True/False)

---

📊 Performance

Metric	Score
Precision (p)	99.55%
Recall (r)	99.10%
F1-score (f1)	99.33%
Epochs Trained	15
Device	CUDA (GPU)

Example from training:

Epoch 15: p=0.9955 r=0.9910 f1=0.9932 time=65.3s

---

🗂 Dataset

• Type: Custom dataset with dish images and dish names.
• Structure:
  • train/ → Training images
  • val/ → Validation images
• Images are labeled according to dish names and mapped to verification labels.

---

🚀 How to Use

from PIL import Image
import torch
from torchvision import transforms
from model import DishVerificationModel  # your model class

# Load model
model = DishVerificationModel()
model.load_state_dict(torch.load("base.pth", map_location="cpu"))
model.eval()

# Preprocess image
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor()
])
img = Image.open("test_dish.jpg")
img_tensor = transform(img).unsqueeze(0)

# Predict
with torch.no_grad():
    output = model(img_tensor)
    prediction = torch.sigmoid(output).item()
    print("Match:", prediction > 0.5)

---

🧪 Example Use Cases

• Restaurant apps verifying menu images
• Food delivery services detecting incorrect uploads
• Dish tagging automation

---

⚠️ Limitations

• May misclassify visually similar dishes.
• Works best with clear, high-quality images.

---

📄 License

This model is released under the Apache 2.0 License.

---

✨ Citation

If you use this model, please cite:

@software{dish_verification_2025,
  title        = {Dish Image Verification Model},
  author       = {Vamsee},
  year         = {2025},
  url          = {https://huggingface.co/vamsee99/food-vs-notfood-model}
}

---