[HN Gopher] HDR-NeRF: High Dynamic Range Neural Radiance Fields
___________________________________________________________________
HDR-NeRF: High Dynamic Range Neural Radiance Fields
Author : PaulHoule
Score : 76 points
Date : 2023-04-26 17:19 UTC (5 hours ago)
(HTM) web link (xhuangcv.github.io)
(TXT) w3m dump (xhuangcv.github.io)
| londons_explore wrote:
| Does this actually add much more than just taking the original
| images with an 'HDR' mode on a camera?
| PaulHoule wrote:
| It lets you synthesize views that you didn't take with the
| camera.
|
| I am into making stereograms and this sort of technology would
| help, for instance maybe the separation of my two cameras is
| more or less than I wanted so I could make up the views I
| really wanted.
| the8472 wrote:
| NeRFs aren't limited to small camera orbits around one
| perspective of a scene either.
|
| https://jonbarron.info/zipnerf/
| jiggawatts wrote:
| They _didn 't_ turn HDR mode on, that's the novel aspect of the
| research. They took SDR photos with different positions _and_
| exposure levels, and then merged them into the NeRF.
|
| Before modern sensors with huge dynamic range for a single
| photo, the "old school" way of getting HDR was to put your
| camera on a tripod, and take 5+ shots with different exposure
| levels. This can then be merged in software, but obviously this
| only works on static scenes and a static camera position.
|
| The HDR-NeRF model lets you move the camera around, you don't
| need a tripod.
|
| PS: Something I find hilarious is that the paper talks about
| HDR and shows "HDR-SDR" comparisons... but both pictures are
| SDR JPG images. It's pathetic how the PC world is
| pathologically unable to uplift imaging standards while the TV
| and Camera industries with far smaller R&D budgets are casually
| sailing past us into the future.
| zamadatix wrote:
| But what advantage does not turning HDR mode on give vs doing
| so is the question. I guess you could be trying to generate a
| NeRF with a really old smart phone?
| pedalpete wrote:
| Use today could be in turning older non-HDR images into HDR
| NeRFs, but also as a foundation that somebody else will
| build something completely new on top of. Who knows what
| that may be.
| cattown wrote:
| How do they decide where to use the little letters in these
| acronyms? Seems like it should either be HDR-NRF or something
| like HiDR-NeRF. Come on folks. Naming consistency, please.
| IanCal wrote:
| More importantly they missed the opportunity to call it NeRF-
| HDR, pronounced Nerf Herder.
| sva_ wrote:
| I personally appreciate abbreviations that easily roll of the
| tongue and seem like words on their own. I believe it's
| important to be able to talk fluently about these things, to
| better relate to and understand them.
| gamegoblin wrote:
| NeRF was already standardized in the literature by people who
| are not these authors [0] and HDR is also already standardized
| by people who are not these authors [1]. Go with the flow
| unless you have a very good reason not to.
|
| HDR is probably HDR and not some other clever name because
| there was no obvious clever name with an easy pronunciation.
| Whereas NRF is just so close to NeRF -- an actual word people
| are familiar with -- it's begging to be pronounced that way.
|
| [0] https://arxiv.org/abs/2003.08934
|
| [1] https://en.wikipedia.org/wiki/High_dynamic_range
| yieldcrv wrote:
| aka its NeRF or nothing
| mgsouth wrote:
| > _HDR is probably HDR and not some other clever name because
| there was no obvious clever name with an easy pronunciation._
|
| Wouldn't matter, people would complain anyway. HDRs gonna H8.
| PaulHoule wrote:
| Doesn't the vowel for e in "neural" match the e in in Nerf?
| yorwba wrote:
| TIL some people pronounce "neural" as /ne@l/ (ner-al)
| instead of /njUr@l/ (new-ral).
| gamegoblin wrote:
| Yes, the vowel matches, but you could imagine the original
| authors had chosen NRF, pronounced "en ar eff". But the
| fact that the e from Neural was there, and the word nerf
| was there, just a perfect alignment.
___________________________________________________________________
(page generated 2023-04-26 23:00 UTC)