4 ms·
That did strike me as weird. Although they say "advanced computer vision software" so maybe they do some workaround and take the position from a camera rather t
by viraptor 7y ago
That did strike me as weird. Although they say "advanced computer vision software" so maybe they do some workaround and take the position from a camera rather than pre-calculations? Can it be the case that a few thousandths of a degree matters here and that's beyond our generalised model?
- TeMPOraL 7y agoI'd imagine that the "few thousandths of a degree" is something you'd get worse with AI than with a mathematical model. AI methods introduce noise; with an explicit model you can keep adding complexity directly, until it gets more accurate than your measuring system. I suspect that the "AI" they mention is mostly photogrammetry - being able to tell the position and orientation of your mirrors from a video stream - in a feedback loop with motor controllers and temperature sensors. Use of cameras would make it close enough to "AI" that the buzzword would stick.
- ovi256 7y ago>mostly photogrammetry - being able to tell the position and orientation of your mirrors from a video stream - in a feedback loop with motor controllers and temperature sensors. This. Modern photogrammetry uses computer vision techniques like deep neural nets. It's an amazing advancement. Imagine this problem: having to place a ground vehicle like a wheeled robot at an arbitrary coordinate in a room. The mirror positioning problem can be reduced to this one. Traditionally, you'd do it by building a very nice wheeled robot with non-slip wheels, wheel rotation angle sensors, fraction of a degree stepper motors, all very nice to work with, precise, and expensive. You could even solve this in an open-loop system with good enough actuators. Today, just stick a camera on the ceiling - or on the robot, if want a more challenging problem - and use CV to move a cheap RC car where you want. The Terminator will jangle and jiggle, instead of being built like a precise Swiss watch. But you'll get ten of them for the same price!
- TeMPOraL 7y ago> This. Modern photogrammetry uses computer vision techniques like deep neural nets. I'm surprised to hear that. What good do DNN do for photogrammetry? Aren't they too unreliable? I only have university course-level exposure to this, most of it spent learning how to compute coordinate systems and transformation matrices from points on a set of pictures taken from different position. I guess DNNs could help with object isolation and labeling (identifying same objects on a set of pictures taken from different perspective)?
- ovi256 7y agoLook for Deep SLAM or visual odometry papers, such as: * https://www.youtube.com/watch?v=Ccj1O7yndIk https://www.youtube.com/watch?v=Ccj1O7yndIk * https://www.youtube.com/watch?v=eOuonMhEsxI https://www.youtube.com/watch?v=eOuonMhEsxI * https://www.groundai.com/project/df-slam-a-deep-learning-enhanced-visual-slam-system-based-on-deep-local-features/ https://www.groundai.com/project/df-slam-a-deep-learning-enh... The Related work section in the DF-SLAM paper is great. As you intuited, DNNs work great for tracking visual landmarks (better than traditional computed features) between frames. They can also estimate depth from a monocular camera, which is very useful. A more out-of-left-field idea is training a CNN to directly regress the camera pose (position and orientation) from an image: PoseNet, https://www.youtube.com/watch?v=u0MVbL_RyPU https://www.youtube.com/watch?v=u0MVbL_RyPU
- marvin 7y agoHow do you train this? By example? Would enjoy seeing an example of this :)