Issue with weights or weights_path/file in ResNet50

General Discussion
,
#1

Hello, this is my very first Tenserflow project, Please help me to solve the issue below the description.

Traceback (most recent call last):
  File "C:\Tensorflow\workspace\training_demo\training2\1.py", line 40, in <module>
    model.add(ResNet50(include_top=False, pooling='avg', weights=resnet_weights_path))
  File "C:\Users\kamil\.conda\envs\tensorflow\lib\site-packages\tensorflow\python\keras\applications\resnet.py", line 457, in ResNet50
    return ResNet(stack_fn, False, True, 'resnet50', include_top, weights,
  File "C:\Users\kamil\.conda\envs\tensorflow\lib\site-packages\tensorflow\python\keras\applications\resnet.py", line 124, in ResNet
    raise ValueError('The `weights` argument should be either '
ValueError: The `weights` argument should be either `None` (random initialization), `imagenet` (pre-training on ImageNet), or the path to the weights file to be loaded.

(tensorflow) C:\Tensorflow\workspace\training_demo\training2>

Code:

image_size = 224

data_generator_with_aug = ImageDataGenerator(preprocessing_function=preprocess_input,

                                             horizontal_flip=True,

                                             width_shift_range=0.2,

                                             height_shift_range=0.2)

train_path='C:/Tensorflow/workspace/training_demo/training2/training'

train_generator = data_generator_with_aug.flow_from_directory(train_path,

                                                     target_size=(image_size, image_size),

                                                     batch_size=24,

                                                     class_mode='categorical')

data_generator_with_no_aug = ImageDataGenerator(preprocessing_function=preprocess_input)

validation_path='C:/Tensorflow/workspace/training_demo/training2/validation'

validation_generator = data_generator_with_no_aug.flow_from_directory(validation_path,

                                                          target_size=(image_size, image_size),

                                                          batch_size=24,

                                                          class_mode='categorical')

resnet_weights_path = 'C:/Tensorflow/workspace/training_demo/training2resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5'#'C:/Tensorflow/workspace/training_demo/training2/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5'

#resnet_weights_path = 'C:/Users/kamil/.keras/models/resnet50_weights_th_dim_ordering_th_kernels_notop.h5'

model = Sequential()

model.add(ResNet50(include_top=False, pooling='avg', weights=resnet_weights_path))

num_classes = 10

model.add(Dense(num_classes, activation='softmax'))

# Say not to train first layer (ResNet) model. It is already trained

model.layers[0].trainable = False

model.compile(optimizer='sgd', loss='categorical_crossentropy', metrics=['accuracy'])

model.fit_generator(train_generator,

steps_per_epoch=3,

epochs=20,

validation_data=validation_generator,

validation_steps=1)

I use my images and my json file with 3 classes.

#3

It seems as though your ResNet50 instantiation inside model.add is not finding your downloaded weights. Additionally, the weight file you’re loading seems to be from this (link) repository. Could you try (any) of the following?

  1. Try re-downloading the weights (sometimes it can become corrupted)
  2. Sometimes a separate Python process may also be running (for example if you opened it in a Jupyter notebook), you may want to ensure only a single notebook is using it at any point in time
  3. Ensure the file path is correct, can you verify that using file explorer that the file exists where it’s supposed to be