Pick a 3D robot perception problem, find a relevant recent paper, build on it, demonstrate the result. Robot hardware is not required, 3D sensor data or simulation is fine. Ongoing research work may be used if it meets the requirements.

Milestone	Due
Teams formed	Fri Sep 25
Proposal, target paper and an evaluation	Fri Oct 23
Checkpoint, reproduced baseline running end to end	Fri Nov 13
Presentation	Dec 14–18, final exam period
Final report and demonstration video	Fri Dec 18

498: a careful reproduction plus one well-motivated extension. 599: a novel component: a new method, a new evaluation, or a negative result established rigorously, in conference format. Same milestones, different rubric.