import gradio as gr
from PIL import Image
import torch
from torchvision import transforms
from transformers import AutoModelForImageClassification, AutoFeatureExtractor

# Cargar el modelo desde Hugging Face Hub
model = AutoModelForImageClassification.from_pretrained("AdrianRevi/Practica1Blindness")
extractor = AutoFeatureExtractor.from_pretrained("AdrianRevi/Practica1Blindness")

# Preprocesamiento
def predict(img: Image.Image):
    inputs = extractor(images=img, return_tensors="pt")
    with torch.no_grad():
        outputs = model(**inputs)
        probs = torch.nn.functional.softmax(outputs.logits, dim=1)[0]
    labels = model.config.id2label
    return {labels[i]: float(probs[i]) for i in range(len(labels))}

# Interfaz Gradio
demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil"),
    outputs=gr.Label(num_top_classes=3),
    examples=["examples/20068.jpg", "examples/20084.jpg"],
    title="Blindness Detection",
    description="Sube una imagen del ojo para detectar el grado de ceguera.",
)

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