app.py

import gradio as gr
from transformers import AutoImageProcessor, AutoModelForObjectDetection
import torch
from PIL import Image, ImageDraw

# Load the model and processor
processor = AutoImageProcessor.from_pretrained("0llheaven/Conditional-detr-finetuned")
model = AutoModelForObjectDetection.from_pretrained("0llheaven/Conditional-detr-finetuned")

def detect_objects(image):
    # Convert image to RGB if it's grayscale
    if image.mode != "RGB":
        image = image.convert("RGB")
    
    # Prepare input for the model
    inputs = processor(images=image, return_tensors="pt")
    outputs = model(**inputs)

    # Filter predictions with confidence greater than 0.5
    target_sizes = torch.tensor([image.size[::-1]])
    results = processor.post_process_object_detection(outputs, target_sizes=target_sizes)

    # Draw bounding boxes around detected objects
    draw = ImageDraw.Draw(image)
    for result in results:
        scores = result["scores"]
        labels = result["labels"]
        boxes = result["boxes"]

        for score, label, box in zip(scores, labels, boxes):
            box = [round(i, 2) for i in box.tolist()]
            label_name = "Pneumonia" if label.item() == 0 else "Other"
            draw.rectangle(box, outline="red", width=3)
            draw.text((box[0], box[1]), f"{label_name}: {round(score.item(), 3)}", fill="red")
    
    return image

# Create the Gradio interface
interface = gr.Interface(
    fn=detect_objects, 
    inputs=gr.Image(type="pil"), 
    outputs=gr.Image(type="pil"),  # Corrected output type
    title="Object Detection with Transformers",
    description="Upload an image to detect objects using a fine-tuned Conditional-DETR model."
)

# Launch the interface
interface.launch()