search

fchollet / deep-learning-with-python-notebooks

Jupyter notebooks for the code samples of the book "Deep Learning with Python"
MIT License
18.17k stars 8.53k forks source link

5.4.3 Visualizing heatmaps of class activation: Dimension Error? #137

Open XZLeo opened 4 years ago

XZLeo commented 4 years ago

This issue occurred when I tried to visualize heatmaps of class activation. First, I copied the sample code on the e-book and got this error:

IndexError                                Traceback (most recent call last)
<ipython-input-4-d07470776a6e> in <module>()
     14 for i in range(512):
     15     
---> 16     conv_layer_output_value[:, :, i] *= pooled_grads_value[i]   
     17     
     18 heatmap = np.mean(conv_layer_output_value, axis=-1)

IndexError: index 14 is out of bounds for axis 2 with size 14

Then I tried to print the shape of conv_layer_output_value.shape, and then I got this: (1, 14, 14, 512) Then I realized the 16th line should be conv_layer_output_value[:, :, :, i] *= pooled_grads_value[i] And this sentence should be added before visualizing the heatmap: heatmap = np.squeeze(heatmap, axis=0)

I've compared other part of my code and sample code, yet found no difference. I think maybe there's a dimension error that should be corrected. If that's the case, should I pull a request?