File size: 2,489 Bytes
4d5be2f de04d07 4d5be2f de04d07 4d5be2f de04d07 4d5be2f 152bd9c de04d07 f2c9943 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 |
import gradio as gr
import torch
import torch.nn as nn
import numpy as np
from PIL import Image
import torchvision.transforms as transforms
# Modelo autoencoder
class Autoencoder(nn.Module):
def __init__(self):
super(Autoencoder, self).__init__()
# Encoder
self.conv1 = nn.Conv2d(1, 32, kernel_size=3, stride=2, padding=1) # 64x64 -> 32x32
self.conv2 = nn.Conv2d(32, 64, kernel_size=3, stride=2, padding=1) # 32x32 -> 16x16
self.conv3 = nn.Conv2d(64, 128, kernel_size=3, stride=2, padding=1) # 16x16 -> 8x8
self.fc1 = nn.Linear(128 * 8 * 8, 32) # Espacio latente
# Decoder
self.fc2 = nn.Linear(32, 128 * 8 * 8)
self.conv4 = nn.ConvTranspose2d(128, 64, kernel_size=3, stride=2, padding=1, output_padding=1) # 8x8 -> 16x16
self.conv5 = nn.ConvTranspose2d(64, 32, kernel_size=3, stride=2, padding=1, output_padding=1) # 16x16 -> 32x32
self.conv6 = nn.ConvTranspose2d(32, 1, kernel_size=3, stride=2, padding=1, output_padding=1) # 32x32 -> 64x64
def encode(self, x):
z = torch.tanh(self.conv1(x))
z = torch.tanh(self.conv2(z))
z = torch.tanh(self.conv3(z))
z = z.view(z.size(0), -1)
z = torch.tanh(self.fc1(z))
return z
def decode(self, x):
z = torch.tanh(self.fc2(x))
z = z.view(z.size(0), 128, 8, 8)
z = torch.tanh(self.conv4(z))
z = torch.tanh(self.conv5(z))
z = torch.sigmoid(self.conv6(z))
return z
def forward(self, x):
return self.decode(self.encode(x))
# Cargar el modelo
model = Autoencoder()
model.load_state_dict(torch.load("autoencoder.pth", map_location=torch.device("cpu")))
model.eval()
# Transformaci贸n de entrada
transform = transforms.Compose([
transforms.Grayscale(),
transforms.Resize((64, 64)),
transforms.ToTensor()
])
# Funci贸n de inferencia
def detectar_anomalia(imagen):
with torch.no_grad():
img_tensor = transform(imagen).unsqueeze(0) # A帽adir batch
reconstruida = model(img_tensor).squeeze(0).squeeze(0)
return reconstruida.numpy() # Convertir a numpy para visualizaci贸n
# Interfaz de Gradio
interface = gr.Interface(
fn=detectar_anomalia,
inputs=gr.Image(type="pil"),
outputs=[gr.Image(type="numpy"), gr.Text()],
title="Detecci贸n de Anomal铆as con Autoencoder",
description="Sube una imagen para detectar anomal铆as usando un autoencoder entrenado."
)
interface.launch()
|