def _load_data(self):
utils.thick_line()
print('Loading data...')
utils.thin_line()
if self.cfg.DATABASE_MODE is not None:
preprocessed_path_ = join(
'../data/{}'.format(self.cfg.DATABASE_MODE),
self.cfg.DATABASE_NAME)
else:
preprocessed_path_ = join(self.cfg.DPP_DATA_PATH,
self.cfg.DATABASE_NAME)
x = utils.load_pkls(preprocessed_path_,
'x_test' + self.append_info,
tl=self.tl_encode,
add_n_batch=1)
y = utils.load_pkls(preprocessed_path_, 'y_test' + self.append_info)
imgs = utils.load_pkls(preprocessed_path_,
'imgs_test' + self.append_info)
utils.thin_line()
print('Data info:')
utils.thin_line()
print('x_test: {}\ny_test: {}\nimgs_test: {}'.format(
x.shape, y.shape, imgs.shape))
return x, y, imgs
def _load_bottleneck_features(self):
"""Load preprocessed bottleneck features."""
utils.thick_line()
print('Loading data...')
utils.thin_line()
x_train = utils.load_pkls(self.preprocessed_path, 'x_train')
x_valid = utils.load_pkls(self.preprocessed_path,
'x_valid',
add_n_batch=1)
y_train = utils.load_pkls(self.preprocessed_path, 'y_train')
y_valid = utils.load_pkls(self.preprocessed_path, 'y_valid')
utils.thin_line()
print('Data info:')
utils.thin_line()
print('x_train: {}\ny_train: {}\nx_valid: {}\ny_valid: {}'.format(
x_train.shape, y_train.shape, x_valid.shape, y_valid.shape))
return x_train, y_train, x_valid, y_valid
def _load_data(self):
"""Load preprocessed data."""
utils.thick_line()
print('Loading data...')
utils.thin_line()
x_train = utils.load_pkls(self.preprocessed_path,
'x_train',
tl=self.tl_encode)
x_valid = utils.load_pkls(self.preprocessed_path,
'x_valid',
tl=self.tl_encode,
add_n_batch=1)
imgs_train = utils.load_pkls(self.preprocessed_path, 'imgs_train')
imgs_valid = utils.load_pkls(self.preprocessed_path, 'imgs_valid')
if imgs_train.shape == x_train.shape:
print('[W] imgs_train.shape == x_train.shape')
del imgs_train
del imgs_valid
gc.collect()
imgs_train = x_train
imgs_valid = x_valid
y_train = utils.load_pkls(self.preprocessed_path, 'y_train')
y_valid = utils.load_pkls(self.preprocessed_path, 'y_valid')
utils.thin_line()
print('Data info:')
utils.thin_line()
print('x_train: {}\ny_train: {}\nx_valid: {}\ny_valid: {}'.format(
x_train.shape, y_train.shape, x_valid.shape, y_valid.shape))
print('imgs_train: {}\nimgs_valid: {}'.format(imgs_train.shape,
imgs_valid.shape))
return x_train, y_train, imgs_train, x_valid, y_valid, imgs_valid
# x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)
# x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)
#
# x_train = x_train.astype('float32')
# x_test = x_test.astype('float32')
# x_train /= 255
# x_test /= 255
# y_train = utils.to_categorical(y_train, num_classes)
# y_test = utils.to_categorical(y_test, num_classes)
# x_train = mutils.load_pkls('../data/preprocessed_data/mnist', 'x_train')
# x_test = mutils.load_pkls('../data/preprocessed_data/mnist', 'x_test')
# y_train = mutils.load_pkls('../data/preprocessed_data/mnist', 'y_train')
# y_test = mutils.load_pkls('../data/preprocessed_data/mnist', 'y_test')
x_train = mutils.load_pkls('../data/preprocessed_data/radical', 'x_train')
x_test = mutils.load_pkls('../data/preprocessed_data/radical', 'x_test')
y_train = mutils.load_pkls('../data/preprocessed_data/radical', 'y_train')
y_test = mutils.load_pkls('../data/preprocessed_data/radical', 'y_test')
#准备自定义的测试样本
#对测试集重新排序并拼接到原来测试集，就构成了新的测试集，每张图片有两个不同数字
idx = list(range(len(x_test)))
np.random.shuffle(idx)
X_test = np.concatenate([x_test, x_test[idx]], 1)
Y_test = np.vstack([y_test.argmax(1), y_test[idx].argmax(1)]).T
X_test = X_test[Y_test[:, 0] != Y_test[:, 1]] #确保两个数字不一样
Y_test = Y_test[Y_test[:, 0] != Y_test[:, 1]]
Y_test.sort(axis=1) #排一下序，因为只比较集合，不比较顺序
# #搭建普通CNN分类模型