Computer vision, from definition to deployment

I'm Patrik Goorts. I build computer vision systems that run in the real world: on phones, on headsets, and on GPUs in the back of a stadium.

That has meant reconstructing people in 3D for holographic calls, interpolating viewpoints between sixteen cameras at a soccer match, and pushing a 4K camera pipeline through an Android phone at 30 fps without a single memory copy. Classical algorithms where they're the right tool, trained networks where they aren't, and CUDA when it has to be fast.

I take problems end to end: define, research, prototype, and deploy