Michael Bloesch

Our new work on imitation learning from the Dyson Robotics Lab, led by Stephen James and @MichaelBloesch. We train in simulation, then adapt to a real robotic task after one demonstration. To be presented at CORL 2018: paper here arxiv.org/abs/1810.03237 youtube.com/watch?v=iCdTo_…

We will present new Dyson Robotics Lab work this week at ECCV. We show that we can learn a non-linear least squares optimiser for dense reconstruction problems which doesn't need hand-designed priors. @ronnieclark__ youtu.be/5bZbMm8UqbA Paper link: arxiv.org/pdf/1809.02966