They mention a couple of features that are not just "long exposure" in the post: * Compensating for moving stars * "Live viewfinder" during exposure * Selectively darkening the sky * Dark current compensation (though that is probably needed for all long-exposure photography...still, not a simple "more exposure" feature)
All easy software implementations. This is not a hardware story.
It’s hardly a secret that innovation in smartphone cameras is mainly in software now. This software camera innovation is one of, or the, main area that phone manufacturers are competing on at the top end of the market, so characterising it as ‘easy’ seems strange.
This is amazing. Now just FYI, for those curious, some recent "old guard" innovations by pro camera makers for astrophotography include: 1. Leveraging the multi-axis image-stabilizing movement available to an in-camera DSLR sensor with GPS for the purpose of tracking the sky during a long-ish exposure to reduce star trails. Ricoh-Pentax Astrotracer. http://www.ricoh-imaging.co.jp/english/photo-life/astro/ 2. Removing…
pixel peeper perhaps, but these images are nice when you look at them on a smartphone screen, print one or view on large screen and the pictures from a mobile phone give a whole different feeling and sense of quality.
What are the possibilities that this AI camera app is not faking the image.
Here is why I ask
- It has gyro - so knows whether we are pointing at sky or not.
- AI checks whether its a clear sky - if yes - post fake image. If no - Dont risk getting caught.
- Time + geo spacing - Gives the angle, position of camera relative to the space above us.
What are the possibilities that this AI camera app is not faking the image. Here is why I ask - It has gyro - so knows whether we are pointing at sky or not. - AI checks whether its a clear sky - if yes - post fake image. If no - Dont risk getting caught. - Time + geo spacing - Gives the angle, position of camera relative to the space above us.
"Faking" isn't thinking about it right. There's no point in going through all the trouble of manually constructing a fake sky photo like that when you can just aggressively train a machine learning model to produce good-looking skies in photographs, which they appear to have done for the purposes of things like adjusting the brightness of the sky in night-time photos. In the end, it's not a question of whether it's "fake", just how much of the resulting photo is the invention of a neural net instead of the result of light hitting the sensor.
In the end photography like this is art, though, so if the person taking the shot is happy with it, then it's fine, probably. Just don't enter it in a competition with rules against retouching...
As long as you manually control the exposure and exposure time, the same can be achieved with an iPhone camera. This is not specific to Pixel.
It seems they have some ML stuff in there for specific features, e.g. sky / land light balance and hot pixel removal (probably similar to how denoising for MC path tracing works). Aside: randomly recognised Ryan Geiss in the credits, he did the Milkdrop plugin for Winamp back in the day, and also some cool tech demos for Nvidia...
There's no machine learning involved for hot pixel removal. The way I have heard it described is that hot pixels stay in the same place across multiple images. One of the processing steps is to figure out the location of the hot pixels by looking at multiple frames. Stars will shift slightly in between the longer exposures where as hot pixels will not.
In that example the Pixel has a 6x longer exposure. While Pixels can currently get better night photos currently, it's simply because the software lets them take minutes-long exposures instead of seconds-long. From a hardware point of view I don't think there's an advantage.
Yep it’s just the settings, hardware is basically the same and can achieve the same result. This is a marketing blog post.
If you ignore all the research and development into creating the software then sure every camera is the same.
This is amazing. Now just FYI, for those curious, some recent "old guard" innovations by pro camera makers for astrophotography include: 1. Leveraging the multi-axis image-stabilizing movement available to an in-camera DSLR sensor with GPS for the purpose of tracking the sky during a long-ish exposure to reduce star trails. Ricoh-Pentax Astrotracer. http://www.ricoh-imaging.co.jp/english/photo-life/astro/ 2. Removing…
I didn't quite like the article. At the begining it shows a good picture of the Milky Way and it says:
> The image has not been retouched or post-processed in any way.
Then the whole article describes how they've automated a full astroprocessing pipeline inside the phone, that makes heavy postprocessing...
This is amazing. Now just FYI, for those curious, some recent "old guard" innovations by pro camera makers for astrophotography include: 1. Leveraging the multi-axis image-stabilizing movement available to an in-camera DSLR sensor with GPS for the purpose of tracking the sky during a long-ish exposure to reduce star trails. Ricoh-Pentax Astrotracer. http://www.ricoh-imaging.co.jp/english/photo-life/astro/ 2. Removing…
I didn't quite like the article. At the begining it shows a good picture of the Milky Way and it says: > The image has not been retouched or post-processed in any way. Then the whole article describes how they've automated a full astroprocessing pipeline inside the phone, that makes heavy postprocessing...
Those lines have been blurring for a good while. Put an Insta filter on your photo? Tut tut, naughty! Your phone's camera app does the same thing for you?* Your phone sure has a good camera!
* Aggressive HDR, sharpening, automatic tint 'fixing', etc. Am I talking about filters or automatic photo processing in flagship phones.
I do get the sentiment in that statement, in that if you take a photo with that phone, you can get the same output without doing much else than pressing the capture button and being still. Very still.
What are the possibilities that this AI camera app is not faking the image. Here is why I ask - It has gyro - so knows whether we are pointing at sky or not. - AI checks whether its a clear sky - if yes - post fake image. If no - Dont risk getting caught. - Time + geo spacing - Gives the angle, position of camera relative to the space above us.
There is no chance this is being faked. I've been playing with the astrophotography mode on a Pixel 4 XL in a remote country location with no cell phone connectivity. A single 4 minute exposure is able to record stars down to about magnitude 9.5. A single exposure can detect the Crab Nebula, or the two brightest satellite galaxies of the Andromeda Galaxy (M 32 and NGC 205). To fake such results without internet connectivity would mean that the camera software would have to have an internal catalog of about 250,000 stars, with accurate colors and coordinates. It would need the size and shape of nebulae, the contours and brightness of the Milky Way, etc. When I add multiple frames, I see fainter stars appear, as they should, so the internal catalog would really need millions of stars, most of which would not show up unless the user carefully aligned multiple individual image and summed them. This is not being faked.