import gradio as gr
from transformers import pipeline

# Load pretrained DETR model for object detection
detection_model = pipeline("object-detection", model="facebook/detr-resnet-50")

# Function to assess vehicle damage
def assess_vehicle_damage(image):
    try:
        # Use the model to predict object locations and labels
        predictions = detection_model(image)

        # Format results to highlight detected objects and potential damage
        report = "🔍 Vehicle Damage Assessment:\n"
        for pred in predictions:
            label = pred['label']
            score = pred['score']
            box = pred['box']
            report += (
                f"- {label} detected with confidence {score:.2f}.\n"
                f"  Location: (X: {box['xmin']:.1f}, Y: {box['ymin']:.1f}, "
                f"Width: {box['xmax'] - box['xmin']:.1f}, Height: {box['ymax'] - box['ymin']:.1f})\n"
            )

        # Add general recommendations based on detected objects
        report += "\n💡 Recommendations:\n"
        if any("car" in pred['label'].lower() for pred in predictions):
            report += "- Inspect detected areas closely for damage severity.\n"
        else:
            report += "- No visible vehicle parts detected. Please upload a clearer image.\n"

        return report
    except Exception as e:
        return f"Error processing the image: {e}"

# Gradio interface
interface = gr.Interface(
    fn=assess_vehicle_damage,
    inputs=gr.Image(type="file", label="Upload Vehicle Image"),
    outputs=gr.Textbox(label="Damage Assessment Report"),
    title="Vehicle Damage Assessor",
    description=(
        "Upload an image of a vehicle to detect damaged parts and get an assessment report. "
        "The app uses advanced AI models to identify objects and predict potential issues."
    ),
    allow_flagging="never"
)

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