File size: 1,675 Bytes
4d5be2f 152bd9c |
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 |
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__()
self.conv1 = nn.Conv2d(1, 32, kernel_size=3, stride=2, padding=1)
self.conv2 = nn.Conv2d(32, 64, kernel_size=3, stride=2, padding=1)
self.fc1 = nn.Linear(64 * 16 * 16, 16)
self.fc2 = nn.Linear(16, 64 * 16 * 16)
self.conv3 = nn.ConvTranspose2d(64, 32, kernel_size=3, stride=2, padding=1, output_padding=1)
self.conv4 = nn.ConvTranspose2d(32, 1, kernel_size=3, stride=2, padding=1, output_padding=1)
def encode(self, x):
z = torch.tanh(self.conv1(x))
z = torch.tanh(self.conv2(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), 64, 16, 16)
z = torch.tanh(self.conv3(z))
z = torch.sigmoid(self.conv4(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()
# Transformación de entrada
transform = transforms.Compose([
transforms.Grayscale(),
transforms.Resize((64, 64)),
transforms.ToTensor()
])
# Función de inferencia
def detectar_anomalia(imagen):
with torch.no_grad():
img_tensor = transform(imagen).unsqueeze(0) # Añadir batch
reconstruida = model(img_tensor).squeeze(0).squeeze(0)
|