They asked
They asked
19 comments
-
-
How could you calculate her height from the photo without her whole body being visible? You can't just "scale" it using the phone as a reference because there is no measurement to scale. Am I missing something or is this bullshit?
-
No, I think it's just bullshit.
-
Maybe she's in frame in the full photo and Xitter is cropping the preview to a square?
-
Right, I guess it's possible. No idea because I haven't visited Xitter in years.
-
-
... You assume standard human anatomical proportions, infer likely age from other visual factors to dial in those proportions more accurately, use other objects in the scene that are nearby that generally have somewhat standardized measurements, like say, countertop height in a bathroom, or, the size of the tiling used in the bath splash wall, or, the common dimensions of a bathroom itself to estimate mirror to rear wall distance.
Heck, maybe even the door itself, those often follow standard dimensions and also dimensions of the sort of basic bevelling type patterns.
If you know even roughly where this person is, say, the USA, well then all that kinda shit is more or less standardized and the same in any even semi modern dwelling.
Its... not very likely that say, she has no legs beneath her knees and is actually in a home built specifically for partial/double amputees.
And you absolutely can use the phone as another relative scale, phones tend to have distinctive features you can use to narrow down which model they could be, phones have published specs that include physical dimensions.
Basically, you just construct a 3d model approximation based off of the 2d image, and fit match.
How do you think facial recognition, gait recognition algos in surveillance cams and otherwise work?
They use a bunch of mapped reference points to estimate things like distance to subject, angular size to actual size, then dial in more more detail like distsnce between eyes, torso length, waist width, limb lengths, etc, to be able to narrow down possible specific people or at least likely group characteristics.
All of these calcs of course have error margins, but more data points and more captures, cross referenced against realistic ranges for each other = less overall error.
All of this kind of stuff is called 'electro optics' when in the military context of say a missile with a visual spectrum camera being used for guidance, been around for 50+ years, generally known as 'computer vision', its used all over the place in many other kinds of common applications, theres even a whole open source version of it.
-
You assume standard human anatomical proportions
If you do that, then it's pretty much just guessing, isn't it? You could also just look up an average height for a woman in her country and (by definition) you would have a big chance of not being too far off. And you're definitely not getting within 1% error margin with that method.
-
-
pretty sure it's bullshit but from this frame the better guess would be to try to calculate length of stretched arms finger to finger, as it's supposed to be equal or very close to your height.
-
... unless you're as lanky as lanky kong, lol.
But yes, human anatomy generally falls within proportional relative ranges, and when they are closer to the high or low end of those ranges, its noticeable, and we tend to use adjectives or phrases that ... we don't all the time realize that that is what they are describing.
When they exceed those general ranges, we have words that are more like medical diagnoses... or sometimes slurs, unfortunately.
Tangential ramble:
OkCupid, way way back before they got bought out by MatchGroup, they once ran and published via blog post a very interesting study on their own user base.
Basically, when it came to faces, there were general trends in the proportionality of different facial feature configurations being seen as broadly more or less attractive by the people that user says they are looking to date.
IE, here is our rough mathematical model of an 'attractive' face vs an 'unattractive' face, which most people broadly agree on, this person is a 3, that one is a 6, this one is a 9, etc.
But there was also something they didn't expect.
Some people, who have particularly uncommon, specific facial feature proportions...
They would not have broad agreement as to who was indicating they were attractive vs unattractive.
Basically, they were if you sorted a reddit/lemmy comment by 'controversial'.
Some group of people very much liked their unique features, others very much disliked it. Lots of very high # ratings vs lots of very low # ratings, as compared to 'typical' people who would have more like a standard bell curve distribution of # scores centered around a mean.
So... to some extent, there is a shared, general consensus of what constitutes an attractive vs unattractive face.
But going along side that, of people with certain rather striking, unique, abnormal features... well, for them, beauty is more in the eye of the specific beholder.
-
-
Fairly easy using the size of the tiles in the shower.
-
-
Should've used her other pictures to guess instead, she probably uploaded a picture with an object with a known height in frame with which you could get an actual size with ±1% variance.
-
The shower tiles.
-
A known height as in you know how far above ground they are. A street sign or something like that. The shower are only individually known, we don't know the total height.
-
-
19 comments
1% random error but the systematic error is probably pretty huge.