Missing How to generate the GIF of filters weights evolving

[roboteck]

Hi Justin Francis
Thanks for this well-written article on visualizing convnets.
In the end, you show a gif of how filters weights evolving. but the code snippet for that is missing
could please add that to the notebook
Thank you

[wagonhelm]

I didn't include how to do it because I did it using imagemagik and the ubuntu terminal which is not compatible with all systems. The original code for creating the gif I lost.

But essentially you have one image that you want to use for the gif. At every say 5 training steps run that image into filter i.e. sess.run(conv1_2, feed_dict={X: someimage.reshape(1,224,224,1)}

In the example I obviously concatenated multiple images together.

Then use scipy to save that image as 'someimage'+str(step) into some directory and create a gif in terminal using

convert -delay 10 -loop 0 ls -v image.gif

I implore you to try and figure it out!

[alexandruvesa]

Hello.
When I try to vizualize CONV 1-1 i get this error. Can you help me please with some suggestions? Thanks a lot, Alex

ValueError Traceback (most recent call last)
C:\Alex\Anaconda\lib\site-packages\tensorflow\python\client\session.py in init(self, fetches, contraction_fn)
299 self._unique_fetches.append(ops.get_default_graph().as_graph_element(
--> 300 fetch, allow_tensor=True, allow_operation=True))
301 except TypeError as e:

C:\Alex\Anaconda\lib\site-packages\tensorflow\python\framework\ops.py in as_graph_element(self, obj, allow_tensor, allow_operation)
3477 with self._lock:
-> 3478 return self._as_graph_element_locked(obj, allow_tensor, allow_operation)
3479

C:\Alex\Anaconda\lib\site-packages\tensorflow\python\framework\ops.py in _as_graph_element_locked(self, obj, allow_tensor, allow_operation)
3561 if obj.graph is not self:
-> 3562 raise ValueError("Operation %s is not an element of this graph." % obj)
3563 return obj

ValueError: Operation name: "init_2"
op: "NoOp"
input: "^conv1_1/weights1_1/Assign"
input: "^conv1_1/biases1_1/Assign"
input: "^conv1_1_1/weights1_1/Assign"
input: "^conv1_1_1/biases1_1/Assign"
is not an element of this graph.

During handling of the above exception, another exception occurred:

ValueError Traceback (most recent call last)
in
1 # Vizualize CONV 1-1
----> 2 test.run(tf.global_variables_initializer())
3 filteredImage = test.run(conv1_1, feed_dict={X: image.reshape(1,224,224,3)})
4 for i in range(32):
5 plt.imshow(filteredImage[:,:,:,i].reshape(224,224))

C:\Alex\Anaconda\lib\site-packages\tensorflow\python\client\session.py in run(self, fetches, feed_dict, options, run_metadata)
927 try:
928 result = self._run(None, fetches, feed_dict, options_ptr,
--> 929 run_metadata_ptr)
930 if run_metadata:
931 proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)

C:\Alex\Anaconda\lib\site-packages\tensorflow\python\client\session.py in _run(self, handle, fetches, feed_dict, options, run_metadata)
1135 # Create a fetch handler to take care of the structure of fetches.
1136 fetch_handler = _FetchHandler(
-> 1137 self._graph, fetches, feed_dict_tensor, feed_handles=feed_handles)
1138
1139 # Run request and get response.

C:\Alex\Anaconda\lib\site-packages\tensorflow\python\client\session.py in init(self, graph, fetches, feeds, feed_handles)
469 """
470 with graph.as_default():
--> 471 self._fetch_mapper = _FetchMapper.for_fetch(fetches)
472 self._fetches = []
473 self._targets = []

C:\Alex\Anaconda\lib\site-packages\tensorflow\python\client\session.py in for_fetch(fetch)
269 if isinstance(fetch, tensor_type):
270 fetches, contraction_fn = fetch_fn(fetch)
--> 271 return _ElementFetchMapper(fetches, contraction_fn)
272 # Did not find anything.
273 raise TypeError('Fetch argument %r has invalid type %r' % (fetch,

C:\Alex\Anaconda\lib\site-packages\tensorflow\python\client\session.py in init(self, fetches, contraction_fn)
305 except ValueError as e:
306 raise ValueError('Fetch argument %r cannot be interpreted as a '
--> 307 'Tensor. (%s)' % (fetch, str(e)))
308 except KeyError as e:
309 raise ValueError('Fetch argument %r cannot be interpreted as a '

ValueError: Fetch argument <tf.Operation 'init_2' type=NoOp> cannot be interpreted as a Tensor. (Operation name: "init_2"
op: "NoOp"
input: "^conv1_1/weights1_1/Assign"
input: "^conv1_1/biases1_1/Assign"
input: "^conv1_1_1/weights1_1/Assign"
input: "^conv1_1_1/biases1_1/Assign"
is not an element of this graph.)