Tech moves fast! Stay ahead of the curve with Techopedia!
Join nearly 200,000 subscribers who receive actionable tech insights from Techopedia.
A capsule network is a kind of shorthand term for a specific kind of neural network pioneered by Stanford scientist Geoffrey Hinton. In the capsule network, specific methodology is applied to image processing to try to affect an understanding of objects from a three-dimensional spectrum.
To understand capsule networks or what Hinton has called the “dynamic routing between capsules” algorithm, it is important to understand convolutional neural networks (CNNs). Convolutional neural networks have done an amazing job of helping computers to assemble features in image processing to understand pictures in some of the same ways that humans do. Complex sets of filtering, pooling and scaling layers help to achieve detailed results. But CNNs are not good at understanding an image from various three-dimensional views.
Hinton's concept is that algorithms such as dynamic routing between capsules can use reverse rendering to break down objects and understand the relationships of their views from various three-dimensional angles. Experts point out that progress in computing power and data storage has made items like capsule networks possible. These interesting ideas form the basis for some current groundbreaking research into more powerful AI.