app.py

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

# Cargar procesador y modelo
processor = DetrImageProcessor.from_pretrained("facebook/detr-resnet-50")
model = DetrForObjectDetection.from_pretrained("facebook/detr-resnet-50")

def detect_objects(image: Image.Image, threshold: float = 0.9):
    if image is None:
        return "Sube una imagen para iniciar."

    # Preprocesar e inferir
    inputs = processor(images=image, return_tensors="pt")
    with torch.no_grad():
        outputs = model(**inputs)

    # (H, W) en enteros; shape (batch, 2)
    h, w = image.size[1], image.size[0]
    target_sizes = torch.tensor([[h, w]], dtype=torch.int64)

    # Post-proceso
    results = processor.post_process_object_detection(
        outputs, target_sizes=target_sizes, threshold=threshold
    )[0]

    # Formatear salida
    lines = []
    for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
        label_name = model.config.id2label[label.item()]
        box_rounded = [round(float(v), 2) for v in box.tolist()]
        lines.append(f"Objeto: {label_name}, Score: {float(score):.2f}, Box: {box_rounded}")

    return "\n".join(lines) if lines else "Sin detecciones (prueba bajando el umbral)."

demo = gr.Interface(
    fn=detect_objects,
    inputs=[gr.Image(type="pil", label="Imagen"), gr.Slider(0, 1, value=0.9, step=0.05, label="Umbral")],
    outputs=gr.Textbox(label="Detecciones"),
    title="Detección de Objetos con Transformers (DETR)",
    description="Sube una imagen y descubre qué objetos puede detectar."
)

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