keras-filter-visualization
keras-filter-visualization copied to clipboard
Visualizing filters by finding images that maximize their outputs
Traceback (most recent call last): File "viz.py", line 118, in vizualizations[i] = visualize_filter(init_img, index, input_placeholder, args.iterations) File "viz.py", line 55, in visualize_filter grads = normalize(grads) File "/Visual/visual/utils.py", line 47, in...
Modified to keras 2.0 and add the tensorflow backend
Hello, I tried to run the sample viz.py code with Theano as backend and got the error message as ``` Using Theano backend. Namespace(img=None, iterations=5, layer='conv5_1', num_filters=16, size=128, weights_path='../vgg16_weights.h5') Traceback...