Capsule networks are a new type of deep neural network. They process visual information in a similar way to 3D computer graphics. They preserve hierarchical relationships between objects, which CNNs fail to do. Although CNNs are still the go-to solutions for most computer vision tasks, they are still facing some challenges. Let's talk about how Capsule networks came to the rescue.
Capsule networks
What convolutional neural networks fail to do
Don't get me wrong—CNNs are great and still the dominant solution for most computer vision problems nowadays. By looking back at what we studied in the previous chapters, CNNs are good at detecting effective features from pixels that are activated by convolutional filters...
