Here is the underlying shape recognition engine behind SceneReader.
In this example, a simple, hand-drawn face activates similar visual ‘memories’ in a database of general shapes.
…an innovative approach to the challenge of machine vision, providing many advantages over existing techniques.
* Unique generalised shape recognition enables a new range of applications.
* A fast algorithm finds simple, meaningful relationships between shapes.
* Shape categories are learned “at first sight” from a single training example.
* Experience grows linearly: new shapes don’t disrupt existing knowledge.
I am still hugely proud of the thinking and effort that went into this, even if no-one else had the vision to invest.
