|
Code
|
Display
|
Definition
|
|
130403
|
Gradient-weighted class activation |
Values are derived using the gradients of a target class flowing into the final convolutional layer of a convolutional neural network to produce a localization map highlighting the important regions in the image for predicting a class. The gradient-weighted class activation map (Grad-CAM) indicates the discriminative image regions used by the network to identify a class. |
|
130402
|
Class activation |
Values are derived using global average pooling in convolutional neural networks to produce a localization map highlighting the important regions in the image for predicting a class. The class activation map (CAM) indicates the discriminative image regions used by the network to identify a class. |
|
130404
|
Saliency |
Values are derived using a single back-propagation pass through a network to produce a localization map highlighting the spatial support of a given class in a given image. |