Spaces:
Sleeping
Sleeping
Download app.py from MARIAFG/Transformers_HG_Master: direct link, hf CLI and curl.
- Browser
- Download file 3.14 kB
-
https://huggingface.co/spaces/MARIAFG/Transformers_HG_Master/resolve/main/app.py
- Command line
-
hf download hf://spaces/MARIAFG/Transformers_HG_Master/app.py
-
curl -L -o app.py https://huggingface.co/spaces/MARIAFG/Transformers_HG_Master/resolve/main/app.py
3.14 kB
| # ============================================================ | |
| # Maestr铆a de Ciencia de Datos | |
| # | |
| # Aprendizaje Autom谩tico Avanzado | |
| # | |
| # Autores: | |
| # - Samir Mateo Sierra Blanco | |
| # - Maria Fernanda Gonz谩lez Acevedo | |
| # ============================================================ | |
| # Se importan las librer铆as necesarias | |
| import torch | |
| import gradio as gr | |
| import spaces | |
| from PIL import Image | |
| from transformers import DetrImageProcessor, DetrForObjectDetection | |
| # Se carga el modelo y el procesador de im谩genes | |
| MODEL_NAME = "facebook/detr-resnet-50" | |
| processor = DetrImageProcessor.from_pretrained(MODEL_NAME) | |
| model = DetrForObjectDetection.from_pretrained(MODEL_NAME) | |
| # Se usa la GPU si est谩 disponible, de lo contrario se usa la CPU | |
| device = torch.device("cuda" if torch.cuda.is_available() else "cpu") | |
| model.to(device) | |
| model.eval() | |
| # Se realiza funci贸n para la detecci贸n de objetos en una imagen | |
| def detect_objects(image: Image.Image): | |
| if image is None: | |
| return "No se recibi贸 ninguna imagen." | |
| # Se asegura que el formato de la imagen sea RGB | |
| image = image.convert("RGB") | |
| # Se realiza el preprocesamiento de la imagen y se pasa al modelo para obtener las predicciones | |
| inputs = processor(images = image, return_tensors = "pt") | |
| # Se mueven los tensores al dispositivo seleccionado | |
| inputs = { | |
| key: value.to(device) | |
| for key, value in inputs.items() | |
| } | |
| # Se realiza la detecci贸n de objetos | |
| with torch.no_grad(): | |
| outputs = model(**inputs) | |
| # Se realiza el postprocesamiento | |
| target_sizes = torch.tensor( | |
| [image.size[::-1]], | |
| device = device | |
| ) | |
| results = processor.post_process_object_detection( | |
| outputs, | |
| target_sizes = target_sizes, | |
| threshold = 0.9 | |
| )[0] | |
| # Se obtienen los resultados y las coordenadas de los cuadros delimitadores | |
| labels = results["labels"] | |
| scores = results["scores"] | |
| boxes = results["boxes"] | |
| detected_objects = [] | |
| for score, label, box in zip(scores, labels, boxes): | |
| label_name = model.config.id2label[label.item()] | |
| x1, y1, x2, y2 = box.tolist() | |
| detected_objects.append( | |
| f"Objeto: {label_name}\n" | |
| f"Score: {score.item():.2f}\n" | |
| f"Box: [{x1:.1f}, {y1:.1f}, {x2:.1f}, {y2:.1f}]" | |
| ) | |
| # Se devuelve el resultado | |
| if not detected_objects: | |
| return "No se detectaron objetos con el nivel de confianza establecido." | |
| return "\n\n".join(detected_objects) | |
| # Se crea la interfaz de usuario con Gradio | |
| def create_interface(): | |
| interface = gr.Interface( | |
| fn = detect_objects, | |
| inputs = gr.Image( | |
| type = "pil", | |
| label = "Sube una imagen" | |
| ), | |
| outputs = gr.Textbox( | |
| label = "Objetos detectados", | |
| lines = 12 | |
| ), | |
| title = "Detecci贸n de objetos con Transformers", | |
| description = "Sube una imagen y descubre qu茅 objetos se pueden detectar." | |
| ) | |
| interface.launch() | |
| # Se ejecuta la interfaz de usuario | |
| if __name__ == "__main__": | |
| create_interface() | |