def main(model, epochs, batch_size, save_intervals, category):
my_model = core.create_model(model, category)
print(
"Python main program for generating images using {} with category {}".
format(model, category))
## preprocess data images if init_train and save the images as pickle file
final_images_stacked = core.load_data(my_model.rows)
my_model.train(data=final_images_stacked,
epochs=epochs,
batch_size=batch_size,
save_intervals=save_intervals,
sample_intervals=save_intervals,
hi_sample_intervals=save_intervals)
action='store_const',
help='do epoch cycle generation')
parser.add_argument('--epoch',
dest='epoch',
default=None,
type=int,
help='specify exact epoch to use')
parser.add_argument('--treshold',
dest='treshold',
default=0.95,
type=float,
help='treshold to select best images in DCGAN')
args = parser.parse_args()
m = core.create_model(args.model)
def main(m, vec):
imgs = m.generate_images(vec)
filename = args.filename.replace('{EP}', str(m.epoch))
filename = filename.replace(
'{D}', "{}{}".format(int(args.use_img), int(args.noise)))
mi, ma = np.min(imgs), np.max(imgs)
imgs = ((imgs - mi) / (ma - mi) * 255).astype(np.uint8)
for i, im in enumerate(imgs):
fn = filename.replace('{N}', str(i))
utils.save_image(im, os.path.join(args.path, fn))
if "VAE" in args.model:
def xover(v1, v2):
v = v1[:]
n = random.randint(1, len(v2) - 2)
v[n:] = v2[n:]
return v
def mutate(v):
n = random.randint(0, len(v) - 1)
v[n] = random.normalvariate(0, 1)
return v
m = core.create_model("DCGAN_1")
vec = np.random.normal(0, 1, (13, m.latent_dim))
while True:
g = m.generate_images(vec)
show_images(g)
n = [int(x) for x in input().split()]
if len(n) != 3: break
# n = [x for x in range(10) if x not in n1]
v = [vec[i, :] for i in n]
for i in range(3):
vec[i] = v[i]
j = 3
for i, k in [(0, 1), (1, 2), (0, 2)]:
vec[j] = xover(v[i], v[k])
vec[j + 1] = xover(v[i], v[k])
import cv2
import os
# 导入项目下的其他需要的包
from core import create_model
from core.predict import predict
from core import create_dataset
# 导入数据集
(train_x, train_y), (test_x, test_y) = create_dataset()
# 导入模型结构
model = create_model()
# 导入模型数据
model.load_weights('./data/weights.h5')
# 用测试集验证模型
out = model.evaluate(test_x, test_y)
# 遍历data/mnist下的图片文件名, 并保存成列表
file_list = os.listdir('./data/mnist/')
# 将列表中文件名排序, 方便之后操作, 不做也可以hhh
file_list.sort(key=lambda x: int(x[:-4]))
# 打印文件名列表的内容, 程序出错的时候可以看看是不是没有找到文件
print('file_list: ', file_list)
# 遍历这个文件列表依次取出里面的文件名来放入模型预测
for file_name in file_list:
# 因为file_name只是文件名, 读取文件的时候需要完整的文件路径, 这里拼接一下文件路径
file_name = os.path.join('./data/mnist/', file_name)
# 打印文件路径方便对照和调试, 不想加可以去掉
print(file_name)
# 读取图片文件
picture = cv2.imread(file_name)