I am following this tutorial.
However, I decided to take the linear layers, And then make them convertible into an image of 1 * 19 * 19. In doing so, I get a bunch of pixels at random places.
Here's what my modified code looks like. To describe what I did, I essentially cut from 0-10 for the labels, Then cut from 10-length of the array for my images. Just so I am separating the labels and jumbled images.
import torch
import torchvision
import torchvision.transforms as transforms
import torch.optim as optim
import torch.nn as nn
import torch.nn.functional as F
import matplotlib.pyplot as plt
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
batch_size = 4
trainset = torchvision.datasets.CIFAR10(root='./data', train=True,
download=True, transform=transform)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size,
shuffle=True)
testset = torchvision.datasets.CIFAR10(root='./data', train=False,
download=True, transform=transform)
testloader = torch.utils.data.DataLoader(testset, batch_size=batch_size,
shuffle=False)
classes = ('plane', 'car', 'bird', 'cat',
'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
class Net(nn.Module):
def __init__(self):
super().__init__()
self.conv1 = nn.Conv2d(3, 6, 5)
self.pool = nn.MaxPool2d(2, 2)
self.conv2 = nn.Conv2d(6, 16, 5)
self.fc1 = nn.Linear(16 * 5 * 5,8 * 7 * 7) # 8 * 7 * 7
self.fc2 = nn.Linear(8 * 7 * 7, 6 * 8 * 8) # 6 * 8 * 8
self.fc3 = nn.Linear(6 * 8 * 8,19 * 19 + 10) # 19 * 19 + 10
def forward(self, x):
x = self.pool(F.relu(self.conv1(x)))
#### IMPORTANT: this here is where I extract the picture of this layer, comparing against the
# regression layers and this here!
self.picx = self.pool(F.relu(self.conv2(x)))
####
x2 = torch.flatten(self.picx, 1) # flatten all dimensions except batch
x2 = F.relu(self.fc1(x2))
x2 = F.relu(self.fc2(x2))
x2 = self.fc3(x2)
return x2
net = Net()
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)
for epoch in range(5): # loop over the dataset multiple times
for i, data in enumerate(trainloader, 0):
# get the inputs; data is a list of [inputs, labels]
inputs, labels = data
# zero the parameter gradients
optimizer.zero_grad()
# forward + backward + optimize
outputs = net(inputs).to(device)
loss = criterion(outputs, labels).to(device)
loss.backward()
optimizer.step()
running_loss = loss.item()
with torch.no_grad():
dataiter = iter(testloader)
images, labels = dataiter.next()
outputs = net.forward(images) ### pic = to the extracted convulotional layer, vs the regression layer.
_, predicted = torch.max(outputs[...,0:10], 1)
print(predicted, labels)
preds = torch.reshape(outputs[...,10:1454], (4,19,19))
plt.imshow(preds[0].detach().numpy())
plt.show()
correct = 0
total = 0
# since we're not training, we don't need to calculate the gradients for our outputs
with torch.no_grad():
for data in testloader:
images, labels = data
# calculate outputs by running images through the network
outputs = net(images)
# the class with the highest energy is what we choose as prediction
_, predicted = torch.max(outputs[...,0:10], 1)
total += labels.size(0)
correct += (predicted == labels).sum().item()
print(f'Accuracy of the network on the 10000 test images: {100 * correct // total} %')
What is this image signifying? And is there any way to get a mask of where the pixels specifically detect the label? In other words, I want to have an image where it draws pixels of where it is seeing a subject, like a dog or a cat.