Faster neural networks straight from JPEG (2018)
1–10 of 108 posts
Re: Faster neural networks straight from JPEG (2018)
#2For those interested, a modern version (vision transformers) was just published this year at CVPR https://openaccess.thecvf.com/content/CVPR2023/html/Park_RGB...
Re: Faster neural networks straight from JPEG (2018)
#3Add "(2018)"
Re: Faster neural networks straight from JPEG (2018)
#4Re: Faster neural networks straight from JPEG (2018)
#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 expectations and information theory.
Re: Faster neural networks straight from JPEG (2018)
#6Add "(2018)"
Back when Uber was thinking it could make self driving taxis
Re: Faster neural networks straight from JPEG (2018)
#7Re: Faster neural networks straight from JPEG (2018)
#8[dead]
Re: Faster neural networks straight from JPEG (2018)
#9Earlier quoted context omitted.
Back when Uber was thinking it could make self driving taxis
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.
Re: Faster neural networks straight from JPEG (2018)
#10Finally 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.