app.py

import gradio as gr
import cv2
import torch
from transformers import DetrImageProcessor, DetrForObjectDetection

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

def detect_objects(video_path):
    # Leer el video
    cap = cv2.VideoCapture(video_path)
    results = []
    
    while cap.isOpened():
        ret, frame = cap.read()
        if not ret:
            break
        
        # Procesar el frame
        inputs = processor(images=frame, return_tensors="pt")
        outputs = model(**inputs)
        
        # Extraer las predicciones
        target_sizes = torch.tensor([frame.shape[:2]])
        results.append(processor.post_process_object_detection(outputs, target_sizes=target_sizes, threshold=0.9)[0])
    
    cap.release()
    return results

# Crear la interfaz de Gradio
iface = gr.Interface(fn=detect_objects, inputs="video", outputs="json")
iface.launch()