[HN Gopher] New algorithm unlocks high-resolution insights for c...
___________________________________________________________________
New algorithm unlocks high-resolution insights for computer vision
Author : zerojames
Score : 39 points
Date : 2024-03-19 20:28 UTC (2 hours ago)
(HTM) web link (news.mit.edu)
(TXT) w3m dump (news.mit.edu)
| TOMDM wrote:
| The papers actual page feels like a clearer explanation to me.
|
| https://mhamilton.net/featup.html
| frozenport wrote:
| Is a learned downsampler a form of inverse crime?
| https://arxiv.org/abs/math-ph/0401050
| fxtentacle wrote:
| What an amazing idea :)
|
| They reproject the input images and run the low-res network
| multiple times. Then they use an approach similar to NeRF to
| merge the knowledge from those reprojected images into a super-
| resolution result.
|
| So in a way, this is quite similar to how modern Pixel phones can
| take a burst of frames and merge them into a final image that has
| a higher resolution than the sensor. Except that they run useful
| AI processing in between and then do the super-resolution merge
| on the results.
| skybrian wrote:
| It's not that clear why they are downsampling and then upsampling
| again. Why not do all the work at the original resolution?
|
| Apparently, the issue is that some vision algorithms only output
| a low-res representation and _that_ needs to be upsampled to
| match the original?
| og_kalu wrote:
| >It's not that clear why they are downsampling and then
| upsampling again. Why not do all the work at the original
| resolution?
|
| For NNs, This is pretty much a compute efficiency thing.
| Working on the original resolution directly is more compute
| intensive.
___________________________________________________________________
(page generated 2024-03-19 23:00 UTC)