import gradio as gr
from transformers import DetrImageProcessor, DetrForObjectDetection
import torch
from PIL import Image, ImageDraw

# Model loading (same as before - with error handling)
try:
    feature_extractor = DetrImageProcessor.from_pretrained("facebook/detr-resnet-50")
    model = DetrForObjectDetection.from_pretrained("facebook/detr-resnet-50", ignore_mismatched_sizes=True)
except Exception as e:  #  Error handling during model loading
    print(f"Error loading model: {e}")  # Log the error so you can see in HF logs
    raise e  # Re-raise for Space to report it

def predict(image):
    inputs = feature_extractor(images=image, return_tensors="pt")
    outputs = model(**inputs)

    target_sizes = torch.tensor([image.size[::-1]])
    results = feature_extractor.post_process_object_detection(outputs, target_sizes=target_sizes, threshold=0.7)[0]

    # Draw bounding boxes on the image
    draw = ImageDraw.Draw(image) # Create a drawing object
    for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
        box = [round(i) for i in box.tolist()]  # Convert to integers for drawing
        draw.rectangle(box, outline="red", width=2) # Outline
        draw.text((box[0], box[1]), model.config.id2label[label.item()], fill="red") # Add a label

    return image  # Return the image with the bounding boxes drawn

# Gradio Interface (updated output type)
iface = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil"),
    outputs=gr.Image(type="pil", label="Detected Potholes (Image)"), # Updated
    title="Pothole Detection POC",
    description="Upload an image to detect potholes."
)

iface.launch()