jacobgil / keras-cam Goto Github PK

Trying to train the network results in the following error:

Traceback (most recent call last):
  File "cam.py", line 60, in <module>
    train(args.dataset_path)
  File "cam.py", line 11, in train
    model = get_model()
  File "~/master/code/keras-cam-master/model.py", line 53, in get_model
    model = VGG16_convolutions()
  File "~/master/code/keras-cam-master/model.py", line 17, in VGG16_convolutions
    model.add(Convolution2D(64, 3, 3, activation='relu', name='conv1_1'))
  File "/Users/username/.virtualenvs/keraspython2/lib/python2.7/site-packages/keras/models.py", line 312, in add
    output_tensor = layer(self.outputs[0])
  File "/Users/username/.virtualenvs/keraspython2/lib/python2.7/site-packages/keras/engine/topology.py", line 487, in __call__
    self.build(input_shapes[0])
  File "/Users/username/.virtualenvs/keraspython2/lib/python2.7/site-packages/keras/layers/convolutional.py", line 410, in build
    self.W = self.init(self.W_shape, name='{}_W'.format(self.name))
  File "/Users/username/.virtualenvs/keraspython2/lib/python2.7/site-packages/keras/initializations.py", line 57, in glorot_uniform
    fan_in, fan_out = get_fans(shape, dim_ordering=dim_ordering)
  File "/Users/username/.virtualenvs/keraspython2/lib/python2.7/site-packages/keras/initializations.py", line 15, in get_fans
    receptive_field_size = np.prod(shape[2:])
  File "/Users/username/.virtualenvs/keraspython2/lib/python2.7/site-packages/numpy/core/fromnumeric.py", line 2539, in prod
    out=out, **kwargs)
  File "/Users/username/.virtualenvs/keraspython2/lib/python2.7/site-packages/numpy/core/_methods.py", line 35, in _prod
    return umr_prod(a, axis, dtype, out, keepdims)
TypeError: unsupported operand type(s) for *: 'NoneType' and 'int'

Running version Keras==1.1.1

Hi Jacob,

Can you please let me what license do you have for your code. I used some bits of your code to build CAM visualizations in a project that will be made public soon, and I need to make sure our code gets the right credit.

Thank you.
George

Rather a question than an issue, but I was wondering whether a Dense layer in between the GAP and the final softmax activation layer could be added.
Thank you.

I had a problem with the approach which is clear in this line:

Hi

When I tried the model by passing an image path I got:

Using Theano backend.
Traceback (most recent call last):
File "cam.py", line 62, in
visualize_class_activation_map(args.model_path, args.image_path, args.output_path)
File "cam.py", line 18, in visualize_class_activation_map
model = load_model(model_path)
File "/home/vyraun/anaconda2/lib/python2.7/site-packages/keras/models.py", line 121, in load_model
f = h5py.File(filepath, mode='r')
File "/home/vyraun/anaconda2/lib/python2.7/site-packages/h5py/_hl/files.py", line 267, in init
name = name.encode(sys.getfilesystemencoding())
AttributeError: 'NoneType' object has no attribute 'encode'

I guess I am not passing the parameter correctly.

I am getting following error while running the command : python cam.py --model_path cam_checkpoint.hdf5 --image_path=image.jpg

ValueError: bad marshal data (unknown type code)

I tried running this in both python versions, updated all the packages but still no success.

Please help.

Hi
I follow your code to get the hidden layer output of an RNN. But it failed when I using K.function([model.layers[0].input], [model.layers[-1].output]), which causes an error:

MissingInputError: ("An input of the graph, used to compute DimShuffle{x,x}(keras_learning_phase), was not provided and not given a value.Use the Theano flag exception_verbosity='high',for more information on this error.", keras_learning_phase)

So is there any trick in your code? Or probably because I load model with model_from_json and model.load_weights?

@jacobgil

I’am using your code for VGG-16 pre-trained model for Keras but with my own weights.

f = h5py.File(weights_path) nb_layers = len(f.attrs["layer_names"]) for k in range(nb_layers): if k >= len(model_MS.layers): # without the last (fully-connected) layers in the savefile break g = f['layer_{}'.format(k)] weights = [g['param_{}'.format(p)] for p in range(g.attrs['nb_params'])] model.layers[k].set_weights(weights) model.layers[k].trainable = False f.close() print

This is the error I'm having:
KeyError: "Can't open attribute (can't locate attribute: 'nb_params')"

g has only this attributes : ['weight_names']

I described my error in this post:
https://stackoverflow.com/questions/58806714/keyerror-cant-open-attribute-cant-locate-attribute-nb-params

Hi, thanks for sharing your good work. I want to adapt your code to make it work for LSTM used on document classification. Any suggestion?

Hello @jacobgil , i want to ask you how to make a bounding box for localization prediction instead of using heatmap?

Can it be implemented into real-time video tracking? if it can, how to implement it? @jacobgil

I downloaded the checkpoint from the mentioned link: https://drive.google.com/open?id=0B1l5JSkBbENBdk95ZW1DOUhqQUE

When i run the command: python cam.py --model_path cam_checkpoint.hdf5 --image_path=image.jpg

I get the following error: File "/home/user/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1570, in _create_c_op
raise ValueError(str(e))
ValueError: Dimension 0 in both shapes must be equal, but are 3 and 64. Shapes are [3,3,5,64] and [64,3,3,3]. for 'Assign' (op: 'Assign') with input shapes: [3,3,5,64], [64,3,3,3].

The full Error is:
Traceback (most recent call last):
File "cam.py", line 61, in
visualize_class_activation_map(args.model_path, args.image_path, args.output_path)
File "cam.py", line 18, in visualize_class_activation_map
model = load_model(model_path)
File "/usr/local/lib/python2.7/dist-packages/keras/models.py", line 246, in load_model
topology.load_weights_from_hdf5_group(f['model_weights'], model.layers)
File "/usr/local/lib/python2.7/dist-packages/keras/engine/topology.py", line 3189, in load_weights_from_hdf5_group
K.batch_set_value(weight_value_tuples)
File "/usr/local/lib/python2.7/dist-packages/keras/backend/tensorflow_backend.py", line 2365, in batch_set_value
assign_op = x.assign(assign_placeholder)
File "/home/user/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/ops/variables.py", line 615, in assign
return state_ops.assign(self._variable, value, use_locking=use_locking)
File "/home/user/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/ops/state_ops.py", line 283, in assign
validate_shape=validate_shape)
File "/home/user/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/ops/gen_state_ops.py", line 60, in assign
use_locking=use_locking, name=name)
File "/home/user/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 787, in _apply_op_helper
op_def=op_def)
File "/home/user/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 3392, in create_op
op_def=op_def)
File "/home/user/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1734, in init
control_input_ops)
File "/home/user/tensorflow/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1570, in _create_c_op
raise ValueError(str(e))
ValueError: Dimension 0 in both shapes must be equal, but are 3 and 64. Shapes are [3,3,5,64] and [64,3,3,3]. for 'Assign' (op: 'Assign') with input shapes: [3,3,5,64], [64,3,3,3].

Any idea why is this happening?

See line 34: https://github.com/jacobgil/keras-cam/blob/master/cam.py

Shouldnt it be class_weights[1, :] instead of class_weights[:, 1]?

If this were a multiclass problem with 3 classes for instance my proposal would take the weight vector associated with that class. The way the code is now it would take the 2nd element in Class 0s weight vector, the 2nd element in Class 1's weight vector, and the 2nd element in Class 3's weight vector.

Am I missing something obvious? I looked at the class weight out put and confirmed that is what your current code is doing on line 34 which I cannot wrap my head around. I may be just confused about how CAM works.