`有人可以帮我解决此代码中的错误吗？ 还有另一个错误，任何尝试运行此代码的人都会看到它会是tensorflow.keras.python.datautils 的导入错误，类似的问题，要解决这个问题，只需进入内置的tensorflow，从tensorflow.keras 中删除.python 部分。 python.datautils 输入文件只是视频中的帧
import cv2
import numpy as np
#import tensorflow as tf
import tensorflow.compat.v1 as tf
import tensornets as nets
from tensornets.utils import load_img
import math
import scipy.io as sio
import time
import random
import itertools
import os
from pprint import pprint
import pickle
import glob
from sklearn.preprocessing import StandardScaler
from sklearn import preprocessing
from keras.models import Sequential
from keras.layers import AveragePooling3D
tf.compat.v1.disable_eager_execution()
batch_size = 32
seg_size = 32
dims = [batch_size, 224, 224, 3]
no_of_videos = 950
global flags
def generate_feat(inputx, model, out_feats, sess):
"""Use the imgload that comes with the model to read the image of each video
Then use the model to process these images to get the output characteristics
Finally save these features as a file in npy format
"""
# Import video list
global flags
seg_name = os.path.basename(flags.input)
vid_names = glob.glob(os.path.join(flags.input, '*'))
res_feats = np.zeros([no_of_videos, seg_size, 2048], np.float32)
# Read seg_size pictures evenly for each video, and process one video as a batch
for idx, vid_n in enumerate(vid_names):
# Read in all the picture names of the video and select seg_size pictures evenly
input_imgs = np.zeros(shape=dims, dtype=np.float32)
vid_idx = int(os.path.basename(vid_n)[:])
fpaths = glob.glob(os.path.join(vid_n, '*'))
frm_len = len(fpaths)
if frm_len == 0:
continue
delta = frm_len / seg_size
idx_list = [int(i*delta) for i in range(seg_size)]
print(idx, vid_n, frm_len, max(idx_list))
# Use load_img to read the images in the list, and model.preprocess for preprocessing
for idx2, idx3 in enumerate(idx_list):
img_path = fpaths[idx3]
img = load_img(img_path, target_size=256, crop_size=224)
input_imgs[idx2,:,:,:] = model.preprocess(img)
feats = sess.run(out_feats, {inputx: input_imgs})
res_feats[vid_idx] = feats
print(idx, 'video has been processed.')
np.save(flags.output, res_feats)
res_feats = res_feats.reshape(1, no_of_videos, 32, 2048, 1)
model = Sequential([AveragePooling3D(pool_size = (1, 32, 1))])
resnet_avg = model.predict(res_feats)
resnet_avg = np.squeeze(resnet_avg)
min_max_scaler = preprocessing.MinMaxScaler()
scaled_resnet = min_max_scaler.fit_transform(resnet_avg.reshape(-1,no_of_videos))
scaled_resnet = scaled_resnet.reshape(no_of_videos,-1)
np.save(flags.scaled_output, scaled_resnet)
if __name__ == "__main__":
global flags
#Directory address to read the frames
tf.app.flags.DEFINE_string('input', '/content/drive/MyDrive/Frames', 'input path')
#Address where npy file has to be saved
tf.app.flags.DEFINE_string('output', '/content/drive/MyDrive/ResNeXt', 'output path')
tf.app.flags.DEFINE_string('scaled_output', '/content/drive/MyDrive/Scaled_ResNeXt', 'scaled output path')
flags = tf.app.flags.FLAGS
# Model file
inputx = tf.placeholder(tf.float32, [None, 224, 224, 3])
model = nets.ResNeXt101c64(inputx, is_training=False)
out_feats = model.get_outputs()[-3]
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
sess.run(tf.global_variables_initializer())
sess.run(model.pretrained())
# Feature
generate_feat(inputx, model, out_feats, sess)
输出
/usr/local/lib/python3.10/dist-packages/tensorflow/python/keras /engine/base_layer_v1.py:1697: UserWarning: `layer.apply` is deprecated and will be removed in a future version. Please use `layer.__call__` method instead.
warnings.warn('`layer.apply` is deprecated and '
---------------------------------------------------------------------------
ModuleNotFoundError Traceback (most recent call last)
<ipython-input-1-d9504b4a8959> in <cell line: 82>()
90 # Model file
91 inputx = tf.placeholder(tf.float32, [None, 224, 224, 3])
---> 92 model = nets.ResNeXt101c64(inputx, is_training=False)
93 out_feats = model.get_outputs()[-3]
94 config = tf.ConfigProto()
17 frames
/usr/local/lib/python3.10/dist-packages/tensornets/utils.py in wrapper(*args, **kwargs)
272 reuse = kwargs.get('reuse', None)
273 with tf.variable_scope(scope, reuse=reuse):
--> 274 x = func(*args, **kwargs)
275 if func.__name__ == 'wrapper':
276 from .middles import direct as p0
/usr/local/lib/python3.10/dist-packages/tensornets/utils.py in wrapper(*args, **kwargs)
354 with arg_scope(layers, outputs_collections=__outputs__):
355 with arg_scopes(layers_args):
--> 356 x = func(*args, **kwargs)
357 x.model_name = func.__name__
358 return x
/usr/local/lib/python3.10/dist-packages/tensornets/resnets.py in resnext101c64(x, is_training, classes, stem, scope, reuse)
227 x = _stack(x, _block3c64, 2048, 3, scope='conv5')
228 return x
--> 229 return resnet(x, False, stack_fn, is_training, classes, stem, scope, reuse)
230
231
/usr/local/lib/python3.10/dist-packages/tensornets/resnets.py in resnet(x, preact, stack_fn, is_training, classes, stem, scope, reuse)
70 x = conv2d(x, 64, 7, stride=2, scope='conv1')
71 else:
---> 72 x = conv(x, 64, 7, stride=2, scope='conv1')
73 x = relu(x, name='conv1/relu')
74 x = pad(x, pad_info(0 if stem else 3, symmetry=not preact),
/usr/local/lib/python3.10/dist-packages/tensornets/layers.py in convbn(*args, **kwargs)
51 scope = kwargs.pop('scope', None)
52 with tf.variable_scope(scope):
---> 53 return batch_norm(conv2d(*args, **kwargs))
54
55
/usr/local/lib/python3.10/dist-packages/tensornets /contrib_framework/arg_scope.py in func_with_args(*args, **kwargs)
180 current_args = current_scope[key_func].copy()
181 current_args.update(kwargs)
--> 182 return func(*args, **current_args)
183
184 _add_op(func)
/usr/local/lib/python3.10/dist-packages/tensornets /contrib_layers/layers.py in batch_norm(inputs, decay, center, scale, epsilon, activation_fn, param_initializers, param_regularizers, updates_collections, is_training, reuse, variables_collections, outputs_collections, trainable, batch_weights, fused, data_format, zero_debias_moving_mean, scope, renorm, renorm_clipping, renorm_decay, adjustment)
627 beta_regularizer = param_regularizers.get('beta')
628 gamma_regularizer = param_regularizers.get('gamma')
--> 629 layer = normalization_layers.BatchNormalization(
630 axis=axis,
631 momentum=decay,
/usr/local/lib/python3.10/dist-packages/tensorflow/python/layers/normalization.py in __getattr__(name)
28 def __getattr__(name):
29 if name in ['BatchNormalization', 'BatchNorm']:
---> 30 return normalization.BatchNormalization
31 elif name in ['batch_normalization', 'batch_norm']:
32 return normalization.batch_normalization
/usr/local/lib/python3.10/dist-packages/tensorflow/python/util/lazy_loader.py in __getattr__(self, item)
64
65 def __getattr__(self, item):
---> 66 module = self._load()
67 return getattr(module, item)
68
/usr/local/lib/python3.10/dist-packages/tensorflow/python/util/lazy_loader.py in _load(self)
47 """Load the module and insert it into the parent's globals."""
48 # Import the target module and insert it into the parent's namespace
---> 49 module = importlib.import_module(self.__name__)
50 self._parent_module_globals[self._local_name] = module
51
/usr/lib/python3.10/importlib/__init__.py in import_module(name, package)
124 break
125 level += 1
--> 126 return _bootstrap._gcd_import(name[level:], package, level)
127
128
/usr/lib/python3.10/importlib/_bootstrap.py in _gcd_import(name, package, level)
/usr/lib/python3.10/importlib/_bootstrap.py in _find_and_load(name, import_)
/usr/lib/python3.10/importlib/_bootstrap.py in _find_and_load_unlocked(name, import_)
/usr/lib/python3.10/importlib/_bootstrap.py in _call_with_frames_removed(f, *args, **kwds)
/usr/lib/python3.10/importlib/_bootstrap.py in _gcd_import(name, package, level)
/usr/lib/python3.10/importlib/_bootstrap.py in _find_and_load(name, import_)
/usr/lib/python3.10/importlib/_bootstrap.py in _find_and_load_unlocked(name, import_)
ModuleNotFoundError: No module named 'keras.legacy_tf_layers'
我尝试查看库以查找 keras.legacy_tf_layers 的导入或使用发生在哪里，但没有找到任何内容。 我尝试更新tensorflow和keras但没有用。