app.py

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

# Load the DETR layout model
model_name = "cmarkea/detr-layout-detection"
processor = DetrImageProcessor.from_pretrained(model_name)
model = DetrForObjectDetection.from_pretrained(model_name)

def detect_layout(image):
    inputs = processor(images=image, return_tensors="pt")
    outputs = model(**inputs)
    target_sizes = torch.tensor([image.size[::-1]])  # H x W
    results = processor.post_process_object_detection(outputs, target_sizes=target_sizes, threshold=0.7)[0]

    draw = ImageDraw.Draw(image)
    labels = []
    for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
        box = [round(i, 2) for i in box.tolist()]
        draw.rectangle(box, outline="red", width=2)
        label_name = model.config.id2label[label.item()]
        draw.text((box[0] + 4, box[1]), f"{label_name} ({round(score.item(), 2)})", fill="red")
        labels.append({"label": label_name, "score": round(score.item(), 2), "box": box})

    return image, labels

iface = gr.Interface(
    fn=detect_layout,
    inputs=gr.Image(type="pil"),
    outputs=[gr.Image(type="pil"), gr.JSON()],
    title="Image to Figma Layers (with DETR)",
    description="Upload a PNG or JPEG UI image to detect editable layers using a layout-aware DETR model."
)

iface.launch()