app.py

import os
import torch
from torchvision.transforms import Compose, Resize, CenterCrop, ToTensor, Normalize
from transformers import ViTImageProcessor
from PIL import Image
import gradio as gr
from huggingface_hub import hf_hub_download

# Repository configuration
REPO_ID = "IFMedTech/Dental_Q"
MODEL_FILENAME = "quantized_model.ptl"

def download_model_from_hub():
    """Download model from private Hugging Face repository"""
    token = os.environ.get("HUGGINGFACE_TOKEN")
    
    if not token:
        raise ValueError(
            "HUGGINGFACE_TOKEN environment variable is required for private repo access. "
            "Please set it in your Space settings under 'Repository secrets'."
        )
    
    try:
        model_path = hf_hub_download(
            repo_id=REPO_ID,
            filename=MODEL_FILENAME,
            token=token
        )
        return model_path
    except Exception as e:
        raise RuntimeError(f"Failed to download model from {REPO_ID}: {str(e)}")

def load_model_and_processor():
    """Load the model and processor"""
    token = os.environ.get("HUGGINGFACE_TOKEN")
    
    # Download and load model
    model_path = download_model_from_hub()
    quantized_model = torch.jit.load(model_path, map_location="cpu")
    quantized_model.eval()
    
    # Load processor from private repo
    processor = ViTImageProcessor.from_pretrained(REPO_ID, token=token)
    
    return quantized_model, processor

# Initialize model and processor
quantized_model, processor = load_model_and_processor()

# Define Inference Preprocessing
size = processor.size['height']
normalize = Normalize(mean=processor.image_mean, std=processor.image_std)
inference_transform = Compose([
    Resize(size),
    CenterCrop(size),
    ToTensor(),
    normalize
])

# Multi-label class names
try:
    label_names = [quantized_model.config.id2label[i] for i in range(len(quantized_model.config.id2label))]
except AttributeError:
    label_names = ["Background", "Caries", "Normal Teeth", "Plaque"]

def preprocess_image(image):
    """Load and preprocess a PIL image."""
    if not isinstance(image, Image.Image):
        image = Image.fromarray(image)
    image = image.convert("RGB")
    return inference_transform(image).unsqueeze(0)

def predict_image(image):
    """Run inference on image and return multi-label predictions."""
    pixel_values = preprocess_image(image)
    
    with torch.no_grad():
        logits = quantized_model(pixel_values)
    
    probs = torch.sigmoid(logits).squeeze(0)
    preds = (probs > 0.5).int().tolist()
    
    detected_conditions = []
    for i, (label, pred) in enumerate(zip(label_names, preds)):
        if pred == 1:
            confidence = probs[i].item()
            detected_conditions.append(f"{label} (confidence: {confidence:.2%})")
    
    # Check for potential Caries
    try:
        caries_index = label_names.index("Caries")
        caries_prob = probs[caries_index].item()
        if 0.3 <= caries_prob < 0.5:
            detected_conditions.append(f"Possible Caries (confidence: {caries_prob:.2%})")
    except ValueError:
        pass
    
    if detected_conditions:
        result = "Detected: " + ", ".join(detected_conditions)
    else:
        result = "No dental issues detected"
    
    return result

# Example images
examples = [
    ["example_image1.jfif"],
    ["example_image2.jfif"],
    ["example_image3.jfif"]
]

# Gradio interface
iface = gr.Interface(
    fn=predict_image,
    inputs=gr.Image(type="pil"),
    outputs="text",
    title="Dental Image Multi-Label Classification",
    description="Upload an image or select from the examples below to predict dental conditions. The model can detect multiple dental issues in a single image.",
    examples=examples
)

if __name__ == "__main__":
    iface.launch()