def predict(latent_dim, height, width, channels):
random_latent_vectors = np.random.normal(size=(100, latent_dim))
dcgan = DCGAN(latent_dim, height, width, channels)
dcgan.load_weights('gan_epoch20.h5')
# dcgan.load_weights('gan_epoch1.h5')
generated_images = dcgan.predict(random_latent_vectors)
for i, generated_image in enumerate(generated_images):
img = image.array_to_img(denormalize(generated_image), scale=False)
img.save(os.path.join('generated', str(i) + '.png'))
if args is None:
exit()
x_train, y_train = create_dataset(
128, 128, nSlices=1000, resize=1,
directory='FluidArt/') # 3 channels = RGB
assert (x_train.shape[0] > 0)
x_train /= 255
# plot results to make sure data looks good!
fig, axs = plt.subplots(4, 4)
for i in range(4):
for j in range(4):
axs[i, j].imshow(x_train[np.random.randint(x_train.shape[0])])
axs[i, j].axis('off')
plt.show()
dcgan = DCGAN(img_rows=x_train[0].shape[0],
img_cols=x_train[0].shape[1],
channels=x_train[0].shape[2],
latent_dim=256,
name='fluid_256_128')
try:
dcgan.load_weights(
generator_file="generator ({}).h5".format(dcgan.name),
discriminator_file="discriminator ({}).h5".format(dcgan.name))
except:
pass
dcgan.train(x_train, epochs=args.epochs, batch_size=32, save_interval=500)
def make_post():
# make sure to load in the correct sized data
dcgan = DCGAN(img_rows = 128,
img_cols = 128,
channels = 3,
latent_dim=256,
name='fluid_256_128')
dcgan.load_weights(generator_file="generator ({}).h5".format(dcgan.name), discriminator_file="discriminator ({}).h5".format(dcgan.name))
# video settings
fps = 30
maxTime = 20 # seconds
frameCount = 0
time = 0
nframes = int( maxTime*fps )
# controls for animation
seed_start = np.random.normal(1, 1, (4, dcgan.latent_dim))
latentSpeed = np.random.normal(3, 1, (4, dcgan.latent_dim))
vary = np.random.normal(1, 1, (4, nframes, dcgan.latent_dim))
# randomize image transformations
#rhue = np.random.random()
rotation = 360 * np.round(np.random.random((4,))*4)/4 # random rotation
flip = np.random.randint(0,4,(4,)) # 0=normal, 1=y-axis, 2=x-axis, 3=transpose
# latent parameter animation
for k in range(4):
time = 0
# for each image in animation
for i in range(nframes):
# change the latent variables
for j in range(dcgan.latent_dim):
vary[k][i][j] = seed_start[k][j] + np.sin( 2*np.pi*(time/maxTime) * latentSpeed[k][j] )
time += 1./fps
imgs = []
for k in range(4):
imgs.append(
dcgan.generator.predict(vary[k])
)
imgs = np.array(imgs)
# create animation
animated_gif = AnimatedGif()
for i in range(nframes):
for k in range(4):
imgs[k,i] = create_mosaic( imgs[k,i], rotation[k], flip[k] )
animated_gif.add(imgs[:,i])
animated_gif.save('artificial_art.mp4',fps=fps)
count = np.loadtxt('count.txt')
with open('hashtags.txt') as fp:
hashtags = fp.readlines()
hashtags = [hashtags[i].strip() for i in range(len(hashtags))]
message = "Automated Artificial Art v 1.{:.1f} - machine hallucinations from an artificial neural network \n \n#".format(count[0])
message = message + " #".join( np.random.choice(hashtags,4,replace=False))
with open('twitter_api_keys.json') as f:
data = json.load(f)
api = twitter.Api(consumer_key=data["consumer_key"],
consumer_secret= data["consumer_secret"],
access_token_key=data["access_token"],
access_token_secret=data["access_secret"])
api.PostUpdate(message,
media="artificial_art.mp4"
)
count += 1
np.savetxt('count.txt',count)
start_seed = None
target_seed = None
steps = 199
linear_step_frames = {}
spherical_step_frames = {}
for n, step_frames in zip(['linear', 'spherical'],
[linear_step_frames, spherical_step_frames]):
for model in models:
gan = DCGAN(name=model, reload=True)
gan.load_weights(250)
if start_seed is None:
start_seed, target_seed = gan.generate_random_noise(2)
for s, val in enumerate([s / steps for s in range(steps + 1)]):
if n == 'spherical':
seed = np.expand_dims(slerp(val, start_seed, target_seed),
axis=0)
elif n == 'linear':
seed = np.expand_dims((val * target_seed), axis=0)
if s not in step_frames:
step_frames[s] = []
os.path.basename(settings["g_weight_path"]).replace(
"generator_", "").replace(".h5", "")) + 1
print("使用するGeneratorの重みは {} ".format(settings["g_weight_path"]))
print()
print("使用するDiscriminatorの重みは {}".format(settings["d_weight_path"]))
# データ・セットの用意
X_test, y_test = load_data(test_path_list)
# データの正規化
X_test = normalize(X_test)
input_shape = X_test[0].shape
# dcganを用意する
dcgan = DCGAN(settings["input_dim"], input_shape)
dcgan.load_weights(d_weight=settings["d_weight_path"],
g_weight=settings["g_weight_path"])
# 保存先rootフォルダの指定
start = time.time()
# AnoGANインスタンスの生成
anogan = ANOGAN(settings["input_dim"], dcgan.g)
anogan.compile(anogan_optim)
with open(save_score_path, "w") as f:
f.write("filenam,anomaly_score" + "\n")
for i, (test_img, test_path) in enumerate(zip(X_test, test_path_list)):
filename = os.path.basename(test_path)[:-4]
loop_start = time.time()
test_img = test_img[np.newaxis, :, :, :]
# parse arguments
args = parse_args()
if args is None:
exit()
x_train, y_train = create_dataset(
128, 128, nSlices=1000, resize=0.5,
directory='Space/Galaxy/') # 3 channels = RGB
assert (x_train.shape[0] > 0)
x_train /= 255
# plot results to make sure data looks good!
fig, axs = plt.subplots(4, 4)
for i in range(4):
for j in range(4):
axs[i, j].imshow(x_train[np.random.randint(x_train.shape[0])])
axs[i, j].axis('off')
plt.show()
dcgan = DCGAN(img_rows=x_train[0].shape[0],
img_cols=x_train[0].shape[1],
channels=x_train[0].shape[2],
latent_dim=128,
name='space_128_128')
dcgan.load_weights(generator_file="generator (space_128_128).h5",
discriminator_file="discriminator (space_128_128).h5")
dcgan.train(x_train, epochs=args.epochs, batch_size=32, save_interval=100)
from keras.optimizers import Adam
from data_loader import load_cucumber
from train import normalize, denormalize
if __name__ == '__main__':
iterations = 100
input_dim = 30
anogan_optim = Adam(lr=0.001, amsgrad=True)
### 0. prepare data
X_train, X_test, y_test = load_cucumber()
X_train = normalize(X_train)
X_test = normalize(X_test)
input_shape = X_train[0].shape
### 1. train generator & discriminator
dcgan = DCGAN(input_dim, input_shape)
dcgan.load_weights('weights/generator_3999.h5', 'weights/discriminator_3999.h5')
for i, test_img in enumerate(X_test):
test_img = test_img[np.newaxis,:,:,:]
anogan = ANOGAN(input_dim, dcgan.g)
anogan.compile(anogan_optim)
anomaly_score, generated_img = anogan.compute_anomaly_score(test_img, iterations)
generated_img = denormalize(generated_img)
imgs = np.concatenate((denormalize(test_img[0]), generated_img[0]), axis=1)
cv2.imwrite('predict' + os.sep + str(int(anomaly_score)) + '_' + str(i) + '.png', imgs)
print(str(i) + ' %.2f'%anomaly_score)
with open('scores.txt', 'a') as f:
f.write(str(anomaly_score) + '\n')