def initialize_fid(train_loader, sample_size=1000):
global base_fid_statistics, inception_model
if inception_model is None:
inception_model = InceptionV3([
InceptionV3.BLOCK_INDEX_BY_DIM[
config.evolution.fitness.fid_dimension]
])
inception_model = tools.cuda(inception_model)
if base_fid_statistics is None:
print("calculate base fid statistics")
# TODO see a better way to load images from the train dataset
train_images = []
for images, _ in train_loader:
train_images += list(images.numpy())
if len(train_images) > sample_size:
train_images = train_images[:sample_size]
break
train_images = np.array(train_images)
base_fid_statistics = fid_score.calculate_activation_statistics(
train_images,
inception_model,
cuda=tools.is_cuda_available(),
dims=config.evolution.fitness.fid_dimension)
inception_model.cpu()
print("completed..")
return base_fid_statistics, inception_model
def getFIDScore(self):
self.Gen_network = self.Gen_network.to(self.device)
self.Gen_network.eval()
noise = torch.randn(1000, 100, 1, 1, device=self.device)
generated_images = self.Gen_network(noise).detach()
self.Gen_network.zero_grad()
self.Gen_network.cpu()
torch.cuda.empty_cache()
# Get FID score for the model:
global base_fid_statistics, inception_model
if (base_fid_statistics is None and inception_model is None):
base_fid_statistics, inception_model = generative_score.initialize_fid(
self.dataloader, sample_size=1000)
inception_model = tools.cuda(inception_model)
m1, s1 = fid_score.calculate_activation_statistics(
generated_images.data.cpu().numpy(),
inception_model,
cuda=tools.is_cuda_available(),
dims=2048)
inception_model.cpu()
m2, s2 = base_fid_statistics
ret = fid_score.calculate_frechet_distance(m1, s1, m2, s2)
torch.cuda.empty_cache()
return ret
def fid_images(generated_images):
global base_fid_statistics, inception_model
inception_model = tools.cuda(inception_model, use_cuda)
start_time = time.time()
m1, s1 = fid_score.calculate_activation_statistics(
generated_images.numpy(), inception_model, cuda=tools.is_cuda_available(use_cuda),
dims=config.evolution.fitness.fid_dimension, batch_size=config.evolution.fitness.fid_batch_size)
print("FID: calc activation --- %s seconds ---" % (time.time() - start_time))
inception_model.cpu()
m2, s2 = base_fid_statistics
ret = fid_score.calculate_frechet_distance(m1, s1, m2, s2)
return ret
def fid_images(generated_images):
global base_fid_statistics, inception_model
inception_model = tools.cuda(inception_model)
m1, s1 = fid_score.calculate_activation_statistics(
generated_images.data.cpu().numpy(),
inception_model,
cuda=tools.is_cuda_available(),
dims=config.evolution.fitness.fid_dimension)
inception_model.cpu()
m2, s2 = base_fid_statistics
ret = fid_score.calculate_frechet_distance(m1, s1, m2, s2)
return ret
def calc_activations(df, image_shape, batch_size):
inception_model = tools.cuda(InceptionV3([InceptionV3.BLOCK_INDEX_BY_DIM[2048]]))
images = get_image_data(df, image_shape).reshape((-1, *image_shape))
print(images.shape)
act = get_activations(images, inception_model, batch_size=batch_size, dims=2048, cuda=tools.is_cuda_available(),
verbose=True)
del inception_model
torch.cuda.empty_cache()
print(act.shape)
return act, images
plt.close()
iters += 1
# Get FID score for the model:
from metrics import generative_score
base_fid_statistics, inception_model = generative_score.initialize_fid(dataloader, sample_size=1000)
from metrics.fid import fid_score
from util import tools
noise = torch.randn(1000, 100, 1, 1, device=device)
netG.eval()
generated_images = netG(noise).detach()
inception_model = tools.cuda(inception_model)
m1, s1 = fid_score.calculate_activation_statistics(
generated_images.data.cpu().numpy(), inception_model, cuda=tools.is_cuda_available(),
dims=2048)
inception_model.cpu()
m2, s2 = base_fid_statistics
ret = fid_score.calculate_frechet_distance(m1, s1, m2, s2)
netG.zero_grad()
print("Fid score is :",ret)
plt.figure(figsize=(10, 5))
plt.title("Generator and Discriminator Loss During Training")
plt.plot(G_losses, label="G")
plt.plot(D_losses, label="D")
plt.xlabel("iterations")
plt.ylabel("Loss")
plt.legend()
