3 ms·
It's fairly easy to dismiss that argument; the cameras would not be difficult to replace if they are the limiting factor. Cameras are cheap.
by jfoster 6y ago
It's fairly easy to dismiss that argument; the cameras would not be difficult to replace if they are the limiting factor. Cameras are cheap.
- sudosysgen 6y agoThe human eye has capabilites that you'd need a camera more expensive than a LiDAR by a good bit too replicate. No rolling shutter, very low delay, continuous signal processing instead of per-frame, maximum resolution around ~70MP eq. in the center, f/2.8 aperture at a full-frame image size, servo driven active rangefinding, continuous cleaning, etc... A camera with those features would be around 3000-15000$, and you'd need two. Also, the human brain can use focus information and stereo phase to deduce 3D structure, which is another ace up its sleeve. It might very well be the case that reproducing this is more expensive in training, processing and material than LiDAR. In fact, I'd say it's very likely - we're talking loops that need to update hundreds of times a second and data rates in the gigabits, which our brains deal with by doing a lot of processing inside the eye and along the way.
- jfoster 6y agoYou don't need human eye equivalents. Human eyes are probably not the ideal cameras for driving.
- sudosysgen 6y agoThis may be the case, you might not have to reproduce everything. But just to give an example that is very relevant to self-driving, getting a camera with similar low-light video performance as the human eye costs over 2000$.
- jfoster 6y agoMy dash cam cost under $200 and seems to have extremely good low light performance.
- sudosysgen 6y agoIn well lit areas, sure. Let's see if your dascham can see an unlit deer at night from a moving, vibrating car (it can't).
- kaba0 6y agoTry (don’t) to drive by looking only at the display of it than..
- Erlich_Bachman 6y ago> Also, the human brain can use focus information and stereo phase to deduce 3D structur Current AI tech can do this too, from multiple images. Karpathy even talked about how they do use that in Teslas.
- sudosysgen 6y agoThe human eye does it in a very different way. It combines both stereo, parallax, eye convergence, and focus. Tesla only does the first.