Robotic vision

I’ve done a lot of work on the commercial development of a software technology, Foveola, which enables machines to locate and identify general shapes within real visual scenes. My wife Kona (who is a real Computer Scientist) and I won a Smart Award in connection with the development of some assistive technology based on this approach (See

Foveola’s unique characteristic is a combination of fast learning and the ability to deal with a huge range of general shapes set against noisy backgrounds. It thus offers the possibility of reliable automated analysis of symbols and icons in complex visual scenes (as indicated on the left).

We then went on to create an application called SceneReader (See This can detect and recognise words in everyday scenes; thus enabling machines to react to signage embedded in their environments. It is not an OCR tool.

Everyone then asks “Can it deal with Chinese signs?” The answer is not yet: we would need about $250,000 of further development funding to make that happen.

I find myself increasingly interested in systems which natural selection has built and which, despite being based on local interactions between simple rules, exhibit emergent ‘behaviours’ which humans often find very hard to model…such as minds themselves. I therefore attempted this paper on the subject of machine consciousness.