805 KiB
805 KiB
!wget https://www.dropbox.com/s/rbajpdlh7efkdo1/male_female_face_images.zip
!unzip -q male_female_face_images.zip!pip install -q --upgrade torch_snippets
from torch_snippets import *
import torchvision
from torchvision import transforms
import torchvision.utils as vutils
import cv2, numpy as np, pandas as pd
device = "cuda" if torch.cuda.is_available() else "cpu"[K |████████████████████████████████| 61kB 6.6MB/s [K |████████████████████████████████| 36.7MB 82kB/s [K |████████████████████████████████| 102kB 13.3MB/s [?25h Building wheel for contextvars (setup.py) ... [?25l[?25hdone
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')!mkdir cropped_faces
images = Glob('/content/females/*.jpg')+Glob('/content/males/*.jpg')
for i in range(len(images)):
img = read(images[i],1)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray, 1.3, 5)
for (x,y,w,h) in faces:
img2 = img[y:(y+h),x:(x+w),:]
cv2.imwrite('cropped_faces/'+str(i)+'.jpg',cv2.cvtColor(img2, cv2.COLOR_RGB2BGR))2020-11-06 18:40:53.399 | INFO | torch_snippets.loader:Glob:181 - 14688 files found at /content/females/*.jpg 2020-11-06 18:40:53.430 | INFO | torch_snippets.loader:Glob:181 - 13948 files found at /content/males/*.jpg
transform=transforms.Compose([
transforms.Resize(64),
transforms.CenterCrop(64),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])class Faces(Dataset):
def __init__(self, folder):
super().__init__()
self.folder = folder
self.images = sorted(Glob(folder))
def __len__(self):
return len(self.images)
def __getitem__(self, ix):
image_path = self.images[ix]
image = Image.open(image_path)
image = transform(image)
return imageds = Faces(folder='cropped_faces/')2020-11-06 18:43:23.108 | INFO | torch_snippets.loader:Glob:181 - 28636 files found at cropped_faces/
dataloader = DataLoader(ds, batch_size=64, shuffle=True, num_workers=8)def weights_init(m):
classname = m.__class__.__name__
if classname.find('Conv') != -1:
nn.init.normal_(m.weight.data, 0.0, 0.02)
elif classname.find('BatchNorm') != -1:
nn.init.normal_(m.weight.data, 1.0, 0.02)
nn.init.constant_(m.bias.data, 0)class Discriminator(nn.Module):
def __init__(self):
super(Discriminator, self).__init__()
self.model = nn.Sequential(
nn.Conv2d(3,64,4,2,1,bias=False),
nn.LeakyReLU(0.2,inplace=True),
nn.Conv2d(64,64*2,4,2,1,bias=False),
nn.BatchNorm2d(64*2),
nn.LeakyReLU(0.2,inplace=True),
nn.Conv2d(64*2,64*4,4,2,1,bias=False),
nn.BatchNorm2d(64*4),
nn.LeakyReLU(0.2,inplace=True),
nn.Conv2d(64*4,64*8,4,2,1,bias=False),
nn.BatchNorm2d(64*8),
nn.LeakyReLU(0.2,inplace=True),
nn.Conv2d(64*8,1,4,1,0,bias=False),
nn.Sigmoid()
)
self.apply(weights_init)
def forward(self, input): return self.model(input)!pip install torch_summary
from torchsummary import summary
discriminator = Discriminator().to(device)
summary(discriminator,torch.zeros(1,3,64,64));Collecting torch_summary Downloading https://files.pythonhosted.org/packages/83/49/f9db57bcad7246591b93519fd8e5166c719548c45945ef7d2fc9fcba46fb/torch_summary-1.4.3-py3-none-any.whl Installing collected packages: torch-summary Successfully installed torch-summary-1.4.3 ========================================================================================== Layer (type:depth-idx) Output Shape Param # ========================================================================================== ├─Sequential: 1-1 [-1, 1, 1, 1] -- | └─Conv2d: 2-1 [-1, 64, 32, 32] 3,072 | └─LeakyReLU: 2-2 [-1, 64, 32, 32] -- | └─Conv2d: 2-3 [-1, 128, 16, 16] 131,072 | └─BatchNorm2d: 2-4 [-1, 128, 16, 16] 256 | └─LeakyReLU: 2-5 [-1, 128, 16, 16] -- | └─Conv2d: 2-6 [-1, 256, 8, 8] 524,288 | └─BatchNorm2d: 2-7 [-1, 256, 8, 8] 512 | └─LeakyReLU: 2-8 [-1, 256, 8, 8] -- | └─Conv2d: 2-9 [-1, 512, 4, 4] 2,097,152 | └─BatchNorm2d: 2-10 [-1, 512, 4, 4] 1,024 | └─LeakyReLU: 2-11 [-1, 512, 4, 4] -- | └─Conv2d: 2-12 [-1, 1, 1, 1] 8,192 | └─Sigmoid: 2-13 [-1, 1, 1, 1] -- ========================================================================================== Total params: 2,765,568 Trainable params: 2,765,568 Non-trainable params: 0 Total mult-adds (M): 106.58 ========================================================================================== Input size (MB): 0.05 Forward/backward pass size (MB): 1.38 Params size (MB): 10.55 Estimated Total Size (MB): 11.97 ==========================================================================================
class Generator(nn.Module):
def __init__(self):
super(Generator,self).__init__()
self.model = nn.Sequential(
nn.ConvTranspose2d(100,64*8,4,1,0,bias=False,),
nn.BatchNorm2d(64*8),
nn.ReLU(True),
nn.ConvTranspose2d(64*8,64*4,4,2,1,bias=False),
nn.BatchNorm2d(64*4),
nn.ReLU(True),
nn.ConvTranspose2d( 64*4,64*2,4,2,1,bias=False),
nn.BatchNorm2d(64*2),
nn.ReLU(True),
nn.ConvTranspose2d( 64*2,64,4,2,1,bias=False),
nn.BatchNorm2d(64),
nn.ReLU(True),
nn.ConvTranspose2d( 64,3,4,2,1,bias=False),
nn.Tanh()
)
self.apply(weights_init)
def forward(self,input): return self.model(input)generator = Generator().to(device)
summary(generator,torch.zeros(1,100,1,1))========================================================================================== Layer (type:depth-idx) Output Shape Param # ========================================================================================== ├─Sequential: 1-1 [-1, 3, 64, 64] -- | └─ConvTranspose2d: 2-1 [-1, 512, 4, 4] 819,200 | └─BatchNorm2d: 2-2 [-1, 512, 4, 4] 1,024 | └─ReLU: 2-3 [-1, 512, 4, 4] -- | └─ConvTranspose2d: 2-4 [-1, 256, 8, 8] 2,097,152 | └─BatchNorm2d: 2-5 [-1, 256, 8, 8] 512 | └─ReLU: 2-6 [-1, 256, 8, 8] -- | └─ConvTranspose2d: 2-7 [-1, 128, 16, 16] 524,288 | └─BatchNorm2d: 2-8 [-1, 128, 16, 16] 256 | └─ReLU: 2-9 [-1, 128, 16, 16] -- | └─ConvTranspose2d: 2-10 [-1, 64, 32, 32] 131,072 | └─BatchNorm2d: 2-11 [-1, 64, 32, 32] 128 | └─ReLU: 2-12 [-1, 64, 32, 32] -- | └─ConvTranspose2d: 2-13 [-1, 3, 64, 64] 3,072 | └─Tanh: 2-14 [-1, 3, 64, 64] -- ========================================================================================== Total params: 3,576,704 Trainable params: 3,576,704 Non-trainable params: 0 Total mult-adds (M): 431.92 ========================================================================================== Input size (MB): 0.00 Forward/backward pass size (MB): 1.97 Params size (MB): 13.64 Estimated Total Size (MB): 15.61 ==========================================================================================
========================================================================================== Layer (type:depth-idx) Output Shape Param # ========================================================================================== ├─Sequential: 1-1 [-1, 3, 64, 64] -- | └─ConvTranspose2d: 2-1 [-1, 512, 4, 4] 819,200 | └─BatchNorm2d: 2-2 [-1, 512, 4, 4] 1,024 | └─ReLU: 2-3 [-1, 512, 4, 4] -- | └─ConvTranspose2d: 2-4 [-1, 256, 8, 8] 2,097,152 | └─BatchNorm2d: 2-5 [-1, 256, 8, 8] 512 | └─ReLU: 2-6 [-1, 256, 8, 8] -- | └─ConvTranspose2d: 2-7 [-1, 128, 16, 16] 524,288 | └─BatchNorm2d: 2-8 [-1, 128, 16, 16] 256 | └─ReLU: 2-9 [-1, 128, 16, 16] -- | └─ConvTranspose2d: 2-10 [-1, 64, 32, 32] 131,072 | └─BatchNorm2d: 2-11 [-1, 64, 32, 32] 128 | └─ReLU: 2-12 [-1, 64, 32, 32] -- | └─ConvTranspose2d: 2-13 [-1, 3, 64, 64] 3,072 | └─Tanh: 2-14 [-1, 3, 64, 64] -- ========================================================================================== Total params: 3,576,704 Trainable params: 3,576,704 Non-trainable params: 0 Total mult-adds (M): 431.92 ========================================================================================== Input size (MB): 0.00 Forward/backward pass size (MB): 1.97 Params size (MB): 13.64 Estimated Total Size (MB): 15.61 ==========================================================================================
def discriminator_train_step(real_data, fake_data):
d_optimizer.zero_grad()
prediction_real = discriminator(real_data)
error_real = loss(prediction_real.squeeze(), torch.ones(len(real_data)).to(device))
error_real.backward()
prediction_fake = discriminator(fake_data)
error_fake = loss(prediction_fake.squeeze(), torch.zeros(len(fake_data)).to(device))
error_fake.backward()
d_optimizer.step()
return error_real + error_fake
def generator_train_step(fake_data):
g_optimizer.zero_grad()
prediction = discriminator(fake_data)
error = loss(prediction.squeeze(), torch.ones(len(real_data)).to(device))
error.backward()
g_optimizer.step()
return errordiscriminator = Discriminator().to(device)
generator = Generator().to(device)
loss = nn.BCELoss()
d_optimizer = optim.Adam(discriminator.parameters(), lr=0.0002, betas=(0.5, 0.999))
g_optimizer = optim.Adam(generator.parameters(), lr=0.0002, betas=(0.5, 0.999))log = Report(25)
for epoch in range(25):
N = len(dataloader)
for i, images in enumerate(dataloader):
real_data = images.to(device)
fake_data = generator(torch.randn(len(real_data), 100, 1, 1).to(device)).to(device)
fake_data = fake_data.detach()
d_loss = discriminator_train_step(real_data, fake_data)
fake_data = generator(torch.randn(len(real_data), 100, 1, 1).to(device)).to(device)
g_loss = generator_train_step(fake_data)
log.record(epoch+(1+i)/N, d_loss=d_loss.item(), g_loss=g_loss.item(), end='\r')
log.report_avgs(epoch+1)
log.plot_epochs(['d_loss','g_loss'])EPOCH: 1.000 d_loss: 0.023 g_loss: 36.309 (45.01s - 1080.18s remaining) EPOCH: 2.000 d_loss: 1.288 g_loss: 7.765 (91.21s - 1048.87s remaining) EPOCH: 3.000 d_loss: 0.725 g_loss: 4.605 (138.52s - 1015.83s remaining) EPOCH: 4.000 d_loss: 0.594 g_loss: 4.297 (187.17s - 982.62s remaining) EPOCH: 5.000 d_loss: 0.575 g_loss: 4.024 (236.63s - 946.51s remaining) EPOCH: 6.000 d_loss: 0.605 g_loss: 3.599 (287.06s - 909.01s remaining) EPOCH: 7.000 d_loss: 0.614 g_loss: 3.401 (337.72s - 868.43s remaining) EPOCH: 8.000 d_loss: 0.646 g_loss: 3.144 (388.28s - 825.09s remaining) EPOCH: 9.000 d_loss: 0.667 g_loss: 3.028 (438.73s - 779.96s remaining) EPOCH: 10.000 d_loss: 0.620 g_loss: 2.907 (489.32s - 733.98s remaining) EPOCH: 11.000 d_loss: 0.661 g_loss: 2.954 (539.77s - 686.98s remaining) EPOCH: 12.000 d_loss: 0.618 g_loss: 2.899 (590.34s - 639.54s remaining) EPOCH: 13.000 d_loss: 0.573 g_loss: 2.901 (640.96s - 591.65s remaining) EPOCH: 14.000 d_loss: 0.613 g_loss: 2.928 (691.42s - 543.26s remaining) EPOCH: 15.000 d_loss: 0.566 g_loss: 3.032 (741.80s - 494.53s remaining) EPOCH: 16.000 d_loss: 0.569 g_loss: 3.048 (792.24s - 445.64s remaining) EPOCH: 17.000 d_loss: 0.496 g_loss: 3.029 (842.80s - 396.61s remaining) EPOCH: 18.000 d_loss: 0.555 g_loss: 3.008 (893.66s - 347.53s remaining) EPOCH: 19.000 d_loss: 0.497 g_loss: 3.129 (944.73s - 298.34s remaining) EPOCH: 20.000 d_loss: 0.554 g_loss: 3.072 (995.37s - 248.84s remaining) EPOCH: 21.000 d_loss: 0.410 g_loss: 3.251 (1045.87s - 199.21s remaining) EPOCH: 22.000 d_loss: 0.480 g_loss: 3.209 (1096.45s - 149.52s remaining) EPOCH: 23.000 d_loss: 0.470 g_loss: 3.323 (1146.98s - 99.74s remaining) EPOCH: 24.000 d_loss: 0.508 g_loss: 3.262 (1197.44s - 49.89s remaining) EPOCH: 25.000 d_loss: 0.414 g_loss: 1.849 (1247.87s - 0.00s remaining)
0%| | 0/25 [00:00<?, ?it/s]/usr/local/lib/python3.6/dist-packages/numpy/core/fromnumeric.py:3335: RuntimeWarning: Mean of empty slice. out=out, **kwargs) /usr/local/lib/python3.6/dist-packages/numpy/core/_methods.py:161: RuntimeWarning: invalid value encountered in double_scalars ret = ret.dtype.type(ret / rcount) 100%|██████████| 25/25 [00:00<00:00, 460.03it/s]
EPOCH: 25.000 d_loss: 0.462 g_loss: 3.340 (1248.02s - 0.00s remaining)
generator.eval()
noise = torch.randn(64, 100, 1, 1, device=device)
sample_images = generator(noise).detach().cpu()
grid = vutils.make_grid(sample_images, nrow=8, normalize=True)
show(grid.cpu().detach().permute(1,2,0), sz=10, title='Generated images')