Hacker News Reader

Comment by trhway | Hacker News Reader

>high-fidelity data from 13 high-resolution cameras

is it me or their example photo doesn't look even like 1 high-resolution camera (and very noisy at that, looks like a pretty high ISO setting). Their example looks similar and i'd say a bit worse than what my Fuji camera did 25 years ago in the dark after i'd adjust dynamic range of the image in GIMP. (and the cameras on the Waymo cars i see around have about the same lens size as my Fuji)

If it is really a showcase of their visual pipeline than it makes sense that they are still so reliant on lidar - which is pretty expensive because of the number of rays you have to support to maintain good situation-awareness and which is still hard and expensive to scale to get good resolution. That also means that their tech can't really be put onto other platforms like drones for example.

View on HN

Absolutely not an expert on this, but I would imagine that they would train the pipeline on a raw-ish image data to have better consistency? I would think that some kind of algorithm "bettering" the image suddenly might cause a discontinuation in the pipeline that is harder to account for versus the ML machinery being trained to do this job as well.

Also, what looks good to us may not be "good" to a machine.

But I'm really just guessing in the dark here.

>they would train the pipeline on a raw-ish image data to have better consistency? I would think that some kind of algorithm "bettering" the image

of course, whatever "bettering" algorithm makes sense, the neural network would probably emerge it inside itself as a result of training. My comment was more about hardware in the sense that "garbage in - garbage out", and so you'd want to have a decent hardware which is available and pretty cheap today while Waymo seems to be using, i'm stating it politely :), very-very-commodity level.

>Also, what looks good to us may not be "good" to a machine.

the kernels of the first layers of visual CNNs (at least the ones i looked at) converge to the Gabor-like kernels what are in the first layers of the human visual cortex. Bunch of years ago, before deep-learning CNN, etc. i saw an article (in neurophysiology field) proving an optimality of such kernels for those visual tasks. Of course things may and probably diverge a lot in the deeper layers, yet deep dreaming, etc. suggests to me that the nature of the divergence may be less severe than expected.

That night image looks very heavily cropped based on the field of view.

Reply on news.ycombinator.com