Capsule networks, a new type of deep neural network, process visual information in much the same way as the brain, which means they can maintain hierarchical relationships. It was conceptualized by Geoffrey Hinton. The motivation behind it to handle the shortcomings of Convolutional Neural Networks (ConvNets).
Comparison of Capsule Networks and CNN
CNN demands lot of data while capsule networks does not require that much data as compare to CNN. CNN are Invariant of Translation which means that they lack to identify the position on an object to another, that is why capsule networks are used because they have the capability to identify the position of a single object.
Usages of Capsule networks
They can be used as -
- Inverse Graphics
- The matrix of weights to represent the relationship
- Dynamic Routing