Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Learning to See in the Dark (2018)

github.com

Post reply on HN
Print view View original

131–140 of 179 posts

Re: Learning to See in the Dark (2018)

#131

Post by barrkel »

anigbrowl wrote: post #128
jameshart wrote: post #120

Earlier quoted context omitted.

Reading a license plate seems like precisely the kind of circumstances where spurious 'plausible interpretation' of limited data can cause trouble. You show in a courtroom a CNN-enhanced low light image of a car and it's there, literally 'clear as day' - the jury will find it pretty compelling. But maybe the data really wasn't there in the original image, and the CNN just filled in some blanks based on previous image…

Of course, but what are the odds that the algorithm just lucked into the correct book title and other cover text? It doesn't have a dictionary or semantic network. You are right that the raw sensor data should always be preserved. But sticking with the license plate example, you could challenge a picture of a single car with a visible license plate far away in a wooded area, but it would be hard to refute a picture o…

Xerox used to replace numbers in documents while copying:

https://www.theregister.co.uk/2013/08/06/xerox_copier_flaw_m...

License plates are an ideal breeding ground for false enhancement owing to standardisation of appearance; an ML algo trained on lots of examples might, without due care, learn to replace as a well-known texture.

Re: Learning to See in the Dark (2018)

#132

Post by mattkrause »

penetrarthur wrote: post #58

I always wondered if you can "trust" an image that has been basically recreated. Could that kind of image be used as an evidence in court?

Xerox copiers had a bug caused by a failed image (re)construction, which caused it to replace similar (but not identical) parts of an image with other pieces of the image.

http://www.dkriesel.com/en/blog/2013/0802_xerox-workcentres_...?

Re: Learning to See in the Dark (2018)

#133

Post by Jnr »

pera wrote: post #127
BubRoss wrote: post #100

Earlier quoted context omitted.

This is completely untrue. What you are describing is usually called automatic tone mapping. This is basically noise reduction and possibly color normalization from brightening a dark image. Them showing their black image as the starting point is silly, because jpg will make a mess of the remaining information. What they should show is the raw image brightened by a straight multiplier to show the noisy version that y…

What jpg? they are using raw data.

Image on Github is JPEG made from RAW. Since RAW file has more dynamic range and contains a lot more information than JPEG you can take that photo in an editor and crank up the brightness. You will get a noisy image but it will be a lot brighter and will probably resemble the image with the high ISO in the middle. Then in an editor you can apply some de-noiser to get results similar to the last one.

So presumably this neural net more or less does it for you.

Re: Learning to See in the Dark (2018)

#134

Post by bialpio »

oconnor663 wrote: post #44
jameshart wrote: post #41

The problem with techniques like this is that they fundamentally amount to ‘making a plausible guess as to what the image would look like’, since essentially they can’t extract information that is simply not there. There is a Shannon entropy limit here. Machine learning is really machine-enhanced educated-guesswork, which has its place but also has its limits.

Counterpoint: The human brain converting a 2D image to a 3D model is educated-guesswork too :)

The courts & hopefully the jurors should be aware that humans are fallible & capable of lying, but may have hard time believing that cameras can lie as well.

Re: Learning to See in the Dark (2018)

#136

Post by gowld »

6gvONxR4sf7o wrote: post #75
jameshart wrote: post #41

The problem with techniques like this is that they fundamentally amount to ‘making a plausible guess as to what the image would look like’, since essentially they can’t extract information that is simply not there. There is a Shannon entropy limit here. Machine learning is really machine-enhanced educated-guesswork, which has its place but also has its limits.

Agreed. It gets to an important point of the purpose of the photo. Photo as a record versus photo as an aesthetic piece. This hurts the photo as a record but improves the photo as an aesthetic piece. This would be a bad addition to a security camera, but perhaps a good addition to an instagram pipeline. There are plenty of other issues there, like is it good/healthy for stuff like instagram to be diverging away from…

Photographs have never been faithful records. The map never the territory. There are always judgement calls. The whole concept of JPEG is to throw away information.

Re: Learning to See in the Dark (2018)

#137

Post by gowld »

bialpio wrote: post #134
oconnor663 wrote: post #44

Earlier quoted context omitted.

Counterpoint: The human brain converting a 2D image to a 3D model is educated-guesswork too :)

The courts & hopefully the jurors should be aware that humans are fallible & capable of lying, but may have hard time believing that cameras can lie as well.

The courts are terrible at that.

Eyewitness testimony is awful but given gold status.

Re: Learning to See in the Dark (2018)

#139

Post by gowld »

ganstyles wrote: post #32
covidacct wrote: post #19

Earlier quoted context omitted.

It's a huge amount of code, hidden behind the tensorflow import statements. It's common to credit GPUs for the rapid spread of deep learning, but good GPUs were available for quite a few years before deep learning really took off. As someone who wrote * a lot* of OpenCL code, including my own python wrappers, I'm fairly certain this code would be thousands of lines without a computation graph framework library. These…

I don't know that I agree with this first statement, but even if I do, everything is abstracted by import statements even outside ML. You say this is a huge amount of code abstracted, but it wouldn't be difficult to reimplement this in numpy and pandas directly without using tensorflow at all. The code would expand a bit, and you'd have to deal directly with backprop and calculating derivatives but it wouldn't expand…

When people say "little code" what they mean an should say is "little customization of the tool".

There's also the issue of how hard it is to select the line of code (API call) that does the job (because the API surface is huge), which is nearly invisible attribute.

Mathematica is famous for being incredibly powerful with low custom coding, but also very hard to find the API call that does the precise thing you need.

Re: Learning to See in the Dark (2018)

#140

Post by SilasX »

jameshart wrote: post #41

The problem with techniques like this is that they fundamentally amount to ‘making a plausible guess as to what the image would look like’, since essentially they can’t extract information that is simply not there. There is a Shannon entropy limit here. Machine learning is really machine-enhanced educated-guesswork, which has its place but also has its limits.

It's a general principle of information theory that, "to make inferences, you have to make assumptions".
Post reply on HN