R&D into next gen AI based vision
R&D into next gen AI based vision, deep learning, multi-camera fusion in latent spaces, embedded algorithms, 3D visual perception, object-centric representation, adaptive neural nets, cognitive architectures, automotive.
I am an entirely obsessed AI / computer vision / cognitive R&D engineer+scientist and entrepreneur since my teens. I started in academic and government research into computer vision back in the mid 80's until 1993 when I started my first company Modular Vision Ltd for developing real-time robust computer vision solutions for automated industrial monitoring and smart manufacturing.
Since 2007 I turned to subcontract research and development across many areas and within several EU collaborative projects, but I found my home since 2010 is automotive computer vision and AI being perhaps one of the earliest proponents of AI and deep neural nets. But since then I realized that without cognition then visual perception is just yet another smart sensor, just now better being based on deep nets. What I realized was missing was obejct- and task-centric visual perception, and that requires cognition. After all, without cognition then vision has no perception since that is what the cognition supplies.
And that cognition is the key that finally opens up the locked door towards robust fully automated driving not only functionally, but in terms of compute resources since cognition supplies task-centric attention. And that means dramatic reduction in compute and memory requirements, but moves away from compute hungry massive SIMD GPU implementations which is a good thing :)
