Live data from Hacker News

Faster neural networks straight from JPEG (2018)

uber.com

11–20 of 108 posts

Re: Faster neural networks straight from JPEG (2018)

#11

Earlier quoted context omitted.

Uber now has a deal with Waymo - so in effect they are still trying to do self-driving taxi's, but now via a complex business relationship.

Ah, just like I "make hamburgers" when I go through the McDonals drive-through, although it's via a business transaction.

God this website is cynical. Licensing technology from another company to commercialize it is completely legitimate.

Re: Faster neural networks straight from JPEG (2018)

#13
post #10

Finally someone who did this, I've always thought this was a low hanging fruit. I wonder if you could make interesting and quickly trained diffusion models with this trick.

Note that it is from 2018. As someone here already mentioned there is a paper that applies the same idea to Vision Transformers published this year [1].

[1] https://openaccess.thecvf.com/content/CVPR2023/papers/Park_R...

Re: Faster neural networks straight from JPEG (2018)

#16
post #5

> Accuracy gains are due primarily to the specific use of a DCT representation, which turns out to work curiously well for image classification. It would seem quantization is a useful tool for any sort of NN-style application. If the expected output is intended to be human-like, why not feed it information that a typical human could not distinguish from a lossless representation? Seems like a simple game of expectati…

That's kind of the key theory behind why JPEG (and other lossy encodings) work at all. A perfect being would see a JPEG next to a PNG or TIFF and find the first repugnantly error-ridden.

But we tend to ignore high-frequency data's specifics most of the time, so it psychologically works.

I often wonder though, what do my cat and dog hear when I'm playing compressed music? Does it sounds like a muddy phone call to them?

Re: Faster neural networks straight from JPEG (2018)

#17
post #10

Finally someone who did this, I've always thought this was a low hanging fruit. I wonder if you could make interesting and quickly trained diffusion models with this trick.

Sorta of? Latent diffusion models operate in a compressed latent space, which is just a richer / learnable representation than DCT.

Re: Faster neural networks straight from JPEG (2018)

#18

For those interested, a modern version (vision transformers) was just published this year at CVPR https://openaccess.thecvf.com/content/CVPR2023/html/Park_RGB...

Thank you. First published 2022-11-29 on arxiv [0] and updated one month ago.

Interesting line: "With these two improvements -- ViT and data augmentation -- we show that our ViT-Ti model achieves up to 39.2% faster training and 17.9% faster inference with no accuracy loss compared to the RGB counterpart."

[0] https://arxiv.org/abs/2211.16421v2

Re: Faster neural networks straight from JPEG (2018)

#19

Earlier quoted context omitted.

Uber now has a deal with Waymo - so in effect they are still trying to do self-driving taxi's, but now via a complex business relationship.

Ah, just like I "make hamburgers" when I go through the McDonals drive-through, although it's via a business transaction.

More like making hamburgers by buying frozen burger patties at the supermarket.

Re: Faster neural networks straight from JPEG (2018)

#20

Earlier quoted context omitted.

Ah, just like I "make hamburgers" when I go through the McDonals drive-through, although it's via a business transaction.

God this website is cynical. Licensing technology from another company to commercialize it is completely legitimate.

That is legitimate, but that's not the point here. The point is Uber's hubris. A hubris very useful in pumping up its stock price ahead of an IPO. If they had quietly planned to license it from the get-go, nobody would have mentioned it.
Post reply on HN