Hana Hussein

Ali Gamal

We used VGG-16 pre-trained model to classify 3 images and interpret the output using different interpretation techniques:

1- First Technique used is "Visualization Using Deconvnets"

We interpreted the model's output on a 'Tiger' image

Then we checked our insights about the filters by:

a) Visualization of Intermediate Activations

b) Visualization of Filters

2- Second Technique used is "Visualization Using Heatmaps"

We interpreted the model's output on a 'Barber Chair' image

Then we checked our insights about the filters by:

a) Visualization Using Deconvnets

2- Second Technique used is "Visualization Using Occlusion Maps"

We interpreted the model's output on a 'French Bulldog' image

Then we checked our insights about the filters by:

a) Visualization Using Heatmaps