What information do you need to build your model of the world?
Inria and IP Paris debut MAGICIAN, using imagined 3D Gaussians to plan robotic exploration paths before physical movement
Drone and toy car tests showed reduced path redundancy.
Positive users praise the MAGICIAN paper for addressing common sense and general intelligence correctly in robot active mapping, while negative users dismiss the work as overhyped and criticize related award decisions.
Most Activity
I think this is my favorite paper this CVPR: Magician.
before they explore in active view selection, they imagine how gaussians and occupancy map would look like and then compute a coverage metric based on that. during planning, they try 10 views like that in 10 steps in a tree search with pruning and get planning for free. they even have real-world experiments with a drone and a toy car.
how are they not an award candidate, it blows my mind.

@chris_j_paxton training data and a lot of failed attempts tbh
you guys made it sound way deeper than it is

@chris_j_paxton Yelp reviews.

I loved this poster because I saw it and I recognised the problem -- we come across it in the context of planning robotic trajectories for inspection camera carriers. What's the shortest time, best path, minimum effort, minimum risk/disruption cost with which I can get a decent walkaround of the object? I loved their solution of constructing essentially the inverse problem.

@chris_j_paxton Now we are addressing priors, common sense, and "general" intelligence in the correct way.

@ftm_guney This year's 3d bitterness might have had a hand for not being in best paper finalist.

@BlissyOnX Yeah thats life for you