今日在使用Tensorflow和Keras做图片分割，在网上找到了这一段代码，自己修改之后保存模型始终保存不上，使用model.save_weight时可以保存，但是使用model.save时就不可以，以下是相关代码：
主函数main：
import numpy as np
import random
import os
from PIL import Image
from keras.utils.np_utils import to_categorical
# Horse dataset, download from http://www.msri.org/people/members/eranb/
def get_horse_generator(horse_path, batch_size=2, input_hw=(256, 256, 3), mask_hw=(256, 256, 20),
preprocess=True, shuffle=True):
def norm(x):
x = x / 128
x -= 1
return x
image_list = ['{:02d}'.format(i) for i in range(49)] # 001 to 250
batch_images = np.empty((batch_size, input_hw[0], input_hw[1], input_hw[2]))
batch_masks = np.empty((batch_size, mask_hw[0], mask_hw[1], mask_hw[2]))
batch_id = 0
while True:
if shuffle:
random.shuffle(image_list)
for image_name in image_list:
try:
image_path = os.path.join(horse_path, 'train\image', image_name + '.jpg')
image = Image.open(image_path)
image = image.resize(input_hw[0:2], Image.NEAREST)
image_np = np.asarray(image, dtype=np.uint8)
if preprocess:
image_np = norm(image_np)
batch_images[batch_id] = image_np
mask_path = os.path.join(horse_path, 'train\label', image_name + '.png')
mask = Image.open(mask_path)
mask = mask.resize(mask_hw[0:2], Image.NEAREST)
mask_np = np.asarray(mask, dtype=np.uint8).copy()
mask_np[mask_np != 0] = 1 # zero, indicating horse.
mask_np = to_categorical(mask_np, num_classes=mask_hw[2])
batch_masks[batch_id] = mask_np
batch_id += 1
if batch_id == batch_size:
batch_id = 0
yield batch_images, batch_masks
except FileNotFoundError:
print('Image not found, Ignore', image_name)
# np.set_printoptions(threshold=np.nan)
# horse_path = 'C:/Users/AlphaCat/Desktop/image_segmentation/weizmann_horse_db'
# horse_gen = get_horse_generator(horse_path, train_or_val='train', batch_size=2, input_hw=(32, 32, 3),
# mask_hw=(32, 32, 2),
# preprocess=True, shuffle=True)
#
# print(next(horse_gen))
# print(next(horse_gen)[0].shape)
# print(next(horse_gen)[1].shape)
fcn处代码：
def get_fcn32s_model(input_shape=(224, 224, 3), class_no=2):
"""
FCN 32 模型
:param input_shape: （输入图片长，输入图片宽，RGB层数）,注意长宽最好是32的倍数
:param class_no: 类别数量
:return: Keras模型
"""
input_tensor = layers.Input(shape=input_shape)
x = layers.ZeroPadding2D(padding=(99, 99))(input_tensor) # Pad 100, 99 + 1 in first layer of vgg
with tf.compat.v1.variable_scope("vgg_encoder"):
encoder = vgg16.VGG16(input_tensor=x, include_top=False, weights='imagenet')
with tf.compat.v1.variable_scope("vgg_decoder"):
x = encoder.get_layer('block5_pool').output
# 卷积做降采用
x = layers.Conv2D(filters=64, kernel_size=(3, 3), activation='relu', padding='same', name='fc6')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters=64, kernel_size=(3, 3), activation='relu', padding='same', name='fc7')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.MaxPooling2D(pool_size=(2, 2), strides=2, padding='same')(x)
x = layers.Dropout(0.2)(x)
x = layers.Conv2D(filters=128, kernel_size=(3, 3), activation='relu', padding='same', name='fc8')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters=128, kernel_size=(3, 3), activation='relu', padding='same', name='fc9')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.MaxPooling2D(pool_size=(2, 2), strides=2, padding='same')(x)
x = layers.Dropout(0.2)(x)
x = layers.Conv2D(filters=256, kernel_size=(3, 3), activation='relu', padding='same', name='fc10')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters=256, kernel_size=(3, 3), activation='relu', padding='same', name='fc11')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters=256, kernel_size=(3, 3), activation='relu', padding='same', name='fc12')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.MaxPooling2D(pool_size=(2, 2), strides=2, padding='same')(x)
x = layers.Dropout(0.2)(x)
x = layers.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name='fc13')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name='fc14')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name='fc15')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.MaxPooling2D(pool_size=(2, 2), strides=2, padding='same')(x)
x = layers.Dropout(0.2)(x)
x = layers.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name='fc16')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name='fc17')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name='fc18')(x)
x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)
x = layers.MaxPooling2D(pool_size=(2, 2), strides=2, padding='same')(x)
x = layers.Dropout(0.2)(x)
# 使用 1x1卷积 做卷积操作，模拟全链接层操作
x = layers.Conv2D(filters=class_no, kernel_size=(1, 1), padding='valid')(x)
# 使用反卷积做Upsampling
x = layers.Conv2DTranspose(filters=class_no, kernel_size=(64, 64), strides=(32, 32), padding='same',
use_bias=False, name='Upsampling1')(x)
# 如果size不够，再做一个Bilinear的Upsampling（通常在图片size不为32的倍数时候需要）
if K.int_shape(x)[1:3] != K.int_shape(input_tensor)[1:3]:
print('Size different, do Bilinear Upsampling')
x = layers.Lambda(lambda x: tf.compat.v1.image.resize_bilinear(x, size=K.int_shape(input_tensor)[1:3]))(x)
# 对输出的每一个像素的各类别（即各通道）的输出使用softmax
x = layers.Activation('relu', name='output')(x)
model = models.Model(inputs=input_tensor, outputs=x)
return model
图片处理的代码：
import numpy as np
import random
import os
from PIL import Image
from keras.utils.np_utils import to_categorical
# Horse dataset, download from http://www.msri.org/people/members/eranb/
def get_horse_generator(horse_path, batch_size=2, input_hw=(256, 256, 3), mask_hw=(256, 256, 20),
preprocess=True, shuffle=True):
def norm(x):
x = x / 128
x -= 1
return x
image_list = ['{:02d}'.format(i) for i in range(49)] # 001 to 250
batch_images = np.empty((batch_size, input_hw[0], input_hw[1], input_hw[2]))
batch_masks = np.empty((batch_size, mask_hw[0], mask_hw[1], mask_hw[2]))
batch_id = 0
while True:
if shuffle:
random.shuffle(image_list)
for image_name in image_list:
try:
image_path = os.path.join(horse_path, 'train\image', image_name + '.jpg')
image = Image.open(image_path)
image = image.resize(input_hw[0:2], Image.NEAREST)
image_np = np.asarray(image, dtype=np.uint8)
if preprocess:
image_np = norm(image_np)
batch_images[batch_id] = image_np
mask_path = os.path.join(horse_path, 'train\label', image_name + '.png')
mask = Image.open(mask_path)
mask = mask.resize(mask_hw[0:2], Image.NEAREST)
mask_np = np.asarray(mask, dtype=np.uint8).copy()
mask_np[mask_np != 0] = 1 # zero, indicating horse.
mask_np = to_categorical(mask_np, num_classes=mask_hw[2])
batch_masks[batch_id] = mask_np
batch_id += 1
if batch_id == batch_size:
batch_id = 0
yield batch_images, batch_masks
except FileNotFoundError:
print('Image not found, Ignore', image_name)
# np.set_printoptions(threshold=np.nan)
# horse_path = 'C:/Users/AlphaCat/Desktop/image_segmentation/weizmann_horse_db'
# horse_gen = get_horse_generator(horse_path, train_or_val='train', batch_size=2, input_hw=(32, 32, 3),
# mask_hw=(32, 32, 2),
# preprocess=True, shuffle=True)
#
# print(next(horse_gen))
# print(next(horse_gen)[0].shape)
# print(next(horse_gen)[1].shape)
报错如下：
Epoch 1/30
2022-04-15 16:51:37.163568: I tensorflow/stream_executor/cuda/cuda_dnn.cc:366] Loaded cuDNN version 8101
2022-04-15 16:51:38.652509: W tensorflow/core/common_runtime/bfc_allocator.cc:275] Allocator (GPU_0_bfc) ran out of memory trying to allocate 3.58GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2022-04-15 16:51:41.660889: W tensorflow/core/common_runtime/bfc_allocator.cc:275] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.16GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2022-04-15 16:51:42.470852: W tensorflow/core/common_runtime/bfc_allocator.cc:275] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.14GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2022-04-15 16:51:42.496813: W tensorflow/core/common_runtime/bfc_allocator.cc:275] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.16GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2022-04-15 16:51:43.549107: W tensorflow/core/common_runtime/bfc_allocator.cc:275] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.80GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
2022-04-15 16:51:43.975971: W tensorflow/core/common_runtime/bfc_allocator.cc:275] Allocator (GPU_0_bfc) ran out of memory trying to allocate 3.58GiB with freed_by_count=0. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.
1/250 [..............................] - ETA: 44:28 - batch: 0.0000e+00 - size: 2.0000 - loss: 1411838.7500 - mean_iou: 0.0000e+00 - acc: 0.53442022-04-15 16:51:44.198713: I tensorflow/core/profiler/lib/profiler_session.cc:110] Profiler session initializing.
2022-04-15 16:51:44.198963: I tensorflow/core/profiler/lib/profiler_session.cc:125] Profiler session started.
2022-04-15 16:51:44.199151: I tensorflow/core/profiler/internal/gpu/cupti_tracer.cc:1630] Profiler found 1 GPUs
2022-04-15 16:51:44.202350: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cupti64_112.dll'; dlerror: cupti64_112.dll not found
2022-04-15 16:51:44.205388: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cupti.dll'; dlerror: cupti.dll not found
2022-04-15 16:51:44.205686: E tensorflow/core/profiler/internal/gpu/cupti_error_manager.cc:135] cuptiGetTimestamp: error 999:
2022-04-15 16:51:44.205951: E tensorflow/core/profiler/internal/gpu/cupti_error_manager.cc:184] cuptiSubscribe: ignored due to a previous error.
2022-04-15 16:51:44.206173: E tensorflow/core/profiler/internal/gpu/cupti_error_manager.cc:457] cuptiGetResultString: ignored due to a previous error.
2022-04-15 16:51:44.206385: E tensorflow/core/profiler/internal/gpu/cupti_tracer.cc:1682] function cupti_interface_->Subscribe( &subscriber_, (CUpti_CallbackFunc)ApiCallback, this)failed with error
2022-04-15 16:51:44.565697: I tensorflow/core/profiler/lib/profiler_session.cc:67] Profiler session collecting data.
2022-04-15 16:51:44.566253: E tensorflow/core/profiler/internal/gpu/cupti_error_manager.cc:140] cuptiFinalize: ignored due to a previous error.
2022-04-15 16:51:44.566520: E tensorflow/core/profiler/internal/gpu/cupti_error_manager.cc:457] cuptiGetResultString: ignored due to a previous error.
2022-04-15 16:51:44.566736: E tensorflow/core/profiler/internal/gpu/cupti_tracer.cc:1773] function cupti_interface_->Finalize()failed with error
2022-04-15 16:51:44.596698: E tensorflow/core/profiler/internal/gpu/cupti_error_manager.cc:133] cuptiGetTimestamp: ignored due to a previous error.
2022-04-15 16:51:44.596994: E tensorflow/core/profiler/internal/gpu/cupti_error_manager.cc:133] cuptiGetTimestamp: ignored due to a previous error.
2022-04-15 16:51:44.597213: I tensorflow/core/profiler/internal/gpu/cupti_collector.cc:526] GpuTracer has collected 0 callback api events and 0 activity events.
2022-04-15 16:51:44.606245: I tensorflow/core/profiler/lib/profiler_session.cc:143] Profiler session tear down.
2022-04-15 16:51:44.618307: I tensorflow/core/profiler/rpc/client/save_profile.cc:136] Creating directory: log\plugins\profile\2022_04_15_08_51_44
2022-04-15 16:51:44.750214: I tensorflow/core/profiler/rpc/client/save_profile.cc:142] Dumped gzipped tool data for trace.json.gz to log\plugins\profile\2022_04_15_08_51_44\LAPTOP-4BUVNCF9.trace.json.gz
2022-04-15 16:51:44.847338: I tensorflow/core/profiler/rpc/client/save_profile.cc:136] Creating directory: log\plugins\profile\2022_04_15_08_51_44
2022-04-15 16:51:44.888927: I tensorflow/core/profiler/rpc/client/save_profile.cc:142] Dumped gzipped tool data for memory_profile.json.gz to log\plugins\profile\2022_04_15_08_51_44\LAPTOP-4BUVNCF9.memory_profile.json.gz
2022-04-15 16:51:45.035196: I tensorflow/core/profiler/rpc/client/capture_profile.cc:251] Creating directory: log\plugins\profile\2022_04_15_08_51_44
Dumped tool data for xplane.pb to log\plugins\profile\2022_04_15_08_51_44\LAPTOP-4BUVNCF9.xplane.pb
Dumped tool data for overview_page.pb to log\plugins\profile\2022_04_15_08_51_44\LAPTOP-4BUVNCF9.overview_page.pb
Dumped tool data for input_pipeline.pb to log\plugins\profile\2022_04_15_08_51_44\LAPTOP-4BUVNCF9.input_pipeline.pb
Dumped tool data for tensorflow_stats.pb to log\plugins\profile\2022_04_15_08_51_44\LAPTOP-4BUVNCF9.tensorflow_stats.pb
Dumped tool data for kernel_stats.pb to log\plugins\profile\2022_04_15_08_51_44\LAPTOP-4BUVNCF9.kernel_stats.pb
250/250 [==============================] - 102s 367ms/step - batch: 124.5000 - size: 2.0000 - loss: 79747.2732 - mean_iou: 0.4251 - acc: 0.6109 - lr: 0.0010
Epoch 2/30
250/250 [==============================] - 91s 365ms/step - batch: 124.5000 - size: 2.0000 - loss: 1044.2393 - mean_iou: 0.4415 - acc: 0.6177 - lr: 0.0010
Epoch 3/30
250/250 [==============================] - ETA: 0s - batch: 124.5000 - size: 2.0000 - loss: 269.6952 - mean_iou: 0.4431 - acc: 0.6153
Epoch 00003: saving model to fcn32s.h5
Traceback (most recent call last):
File "E:\anaconda3\envs\hr\lib\site-packages\IPython\core\interactiveshell.py", line 3457, in run_code
exec(code_obj, self.user_global_ns, self.user_ns)
File "<ipython-input-2-2c9bf53164ed>", line 1, in <module>
runfile('E:/pycharm数据/AI/image_segmentation-master/main.py', wdir='E:/pycharm数据/AI/image_segmentation-master')
File "E:\PyCharm 2021.3\plugins\python\helpers\pydev\_pydev_bundle\pydev_umd.py", line 198, in runfile
pydev_imports.execfile(filename, global_vars, local_vars) # execute the script
File "E:\PyCharm 2021.3\plugins\python\helpers\pydev\_pydev_imps\_pydev_execfile.py", line 18, in execfile
exec(compile(contents+"\n", file, 'exec'), glob, loc)
File "E:/pycharm数据/AI/image_segmentation-master/main.py", line 39, in <module>
model.fit(
File "E:\anaconda3\envs\hr\lib\site-packages\keras\engine\training_v1.py", line 777, in fit
return func.fit(
File "E:\anaconda3\envs\hr\lib\site-packages\keras\engine\training_generator_v1.py", line 570, in fit
return fit_generator(
File "E:\anaconda3\envs\hr\lib\site-packages\keras\engine\training_generator_v1.py", line 315, in model_iteration
callbacks.on_epoch_end(epoch, epoch_logs)
File "E:\anaconda3\envs\hr\lib\site-packages\keras\callbacks.py", line 414, in on_epoch_end
callback.on_epoch_end(epoch, logs)
File "E:\anaconda3\envs\hr\lib\site-packages\keras\callbacks.py", line 1376, in on_epoch_end
self._save_model(epoch=epoch, batch=None, logs=logs)
File "E:\anaconda3\envs\hr\lib\site-packages\keras\callbacks.py", line 1442, in _save_model
self.model.save(filepath, overwrite=True, options=self._options)
File "E:\anaconda3\envs\hr\lib\site-packages\keras\utils\traceback_utils.py", line 67, in error_handler
raise e.with_traceback(filtered_tb) from None
File "E:\anaconda3\envs\hr\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 230, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 205, in _deepcopy_list
append(deepcopy(a, memo))
File "E:\anaconda3\envs\hr\lib\copy.py", line 172, in deepcopy
y = _reconstruct(x, memo, *rv)
File "E:\anaconda3\envs\hr\lib\copy.py", line 296, in _reconstruct
value = deepcopy(value, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 230, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 210, in _deepcopy_tuple
y = [deepcopy(a, memo) for a in x]
File "E:\anaconda3\envs\hr\lib\copy.py", line 210, in <listcomp>
y = [deepcopy(a, memo) for a in x]
File "E:\anaconda3\envs\hr\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 210, in _deepcopy_tuple
y = [deepcopy(a, memo) for a in x]
File "E:\anaconda3\envs\hr\lib\copy.py", line 210, in <listcomp>
y = [deepcopy(a, memo) for a in x]
File "E:\anaconda3\envs\hr\lib\copy.py", line 172, in deepcopy
y = _reconstruct(x, memo, *rv)
File "E:\anaconda3\envs\hr\lib\copy.py", line 270, in _reconstruct
state = deepcopy(state, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 230, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 172, in deepcopy
y = _reconstruct(x, memo, *rv)
File "E:\anaconda3\envs\hr\lib\copy.py", line 270, in _reconstruct
state = deepcopy(state, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 230, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 172, in deepcopy
y = _reconstruct(x, memo, *rv)
File "E:\anaconda3\envs\hr\lib\copy.py", line 270, in _reconstruct
state = deepcopy(state, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 146, in deepcopy
y = copier(x, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 230, in _deepcopy_dict
y[deepcopy(key, memo)] = deepcopy(value, memo)
File "E:\anaconda3\envs\hr\lib\copy.py", line 161, in deepcopy
rv = reductor(4)
TypeError: cannot pickle '_thread.RLock' object
去别处找寻答案时，说可能是因为输入格式的问题，但是并没有找到我这儿对应的代码，希望给各位能给出点解答！谢谢大家了！