Pixel Scaling.

[Vzearr]

Pixel Scaling and it's differing results haunts me. However, one thing I know is that we shouldn't pick a small object to pixel scale as it almost always leads to inflated numbers.

Image link for all 3 cases.

Case 1:
Mikey is 162 cm tall. His height in pixels is 800 pixels.

Now during my short time as a calc group member I noticed lots of people using head length to calculate craters instead of bodies, which was odd to me.

Mikey's head height would be 23.2 cm according to this site which is widely accepted and used.

Mikey's head height here is 108 pixels or 23.2 cm. According to this, Mikey's new height is 172 Cm. Now that's an extra 10cm added to his height.

Now lets get smaller.

Mikey's eye length here is 7 pixels or 2.3 cm. According to this, Mikey's new height is 262 Cm. Now that's an extra 100 cm added to his height!!

Case 2:

Peter's height is 183 cm. tall. His height is pixels is 674.

Peter's head height would be 22.9 cm according to this site which is widely accepted and used.

Peter's head height here is 65 pixels or 22.9 cm. According to this, Peter's new height is 237 Cm. Now that's an extra 54 cm added to his height.

Case 3:
Gongseop is 178 cm tall. His height in pixels is 139 pixels.

Gongseop's head height would be 22.3 cm according to this site which is widely accepted and used.

Gongseop's head height here is 16 pixels or 22.3 cm. According to this, Gongseop's new height is 193 Cm. Now that's an extra 15 cm added to his height.

----------
Now, to show some Calculations that are inflated by this.

Case 1:

Case 2:

Case 3:

I believe. Consistently, that the smaller the reference limb, the more inflated results become, therefore I would like to propose a rule. Something along the lines of, the larger the reference the better.

​

 

[KLOL506]

@Agnaa @Mr. Bambu @DontTalkDT @Executor_N0 @CloverDragon03 @Dark-Carioca @Therefir @DMUA @Migue79 @Psychomaster35 @SunDaGamer @Damage3245 @Armorchompy @Dalesean027 @Flashlight237 @SeijiSetto @DemiiPowa

 

[Damage3245]

Consistently, that the smaller the reference limb, the more inflated results become, therefore I would like to propose a rule. Something along the lines of, the larger the reference the better.

As a guideline for pixelscaling, this sounds like a fair improvement to me. I've seen some calcs where the user tries to pixelscale off of a reference object that is only a single pixel tall/wide. I don't think we can put much stock in calcs like that.

 

[Chariot190]

Pixel Scaling and it's differing results haunts me. However, one thing I know is that we shouldn't pick a small object to pixel scale as it almost always leads to inflated numbers.

Asked three mods, got perm, can show if asked.
I ran into that problem just the other day ironically.

But there's an easy solution. Just don't do it.
It's only an issue due to ratios/resolution (assuming it isn't art inconsistency because that can in fact lead to off values, simply because the artist didn't draw it perfectly proportionately accurate, which happens a lot) or angles and depth which is why we sometimes use trigonomics to get true sizes and distances if things slanted, at angles, skewed, etc, which obviously isn't the pixel scalings fault, but rather just how depth and angles work, which can be accounted for.

Say we have a 1000px image, and the software, having to use and count in pixels, says "yeah this 50cm object is like 2px", you're forced to then treat every px, using that thing as a point of reference, as being equal to 25cm. And sometimes that might not be the case and it's obvious, that's just a limitation via art, software, and so forth.
But in that scenario just figure out a different path, it's easy to solve and avoid, and the margin of error decreases very, very, very, quickly the more higher resolution/higher overall px count an image is in response to the proportionately small thing being scaled (In most cases, in some odd cases for like live action, camera lenses and filters might skew things ever so slightly but that's something else entirely).
And then there's cases where where a small object, being small in pixels, is proportionate, accurate and scales properly in respect to the other aspects of the image, which is also quite common so while this can be an issue, it isn't always an going to be one.

Don't think we really need to set solid ground rule, mostly because it's completely impossible and it depends entirely on the image, panel, screenshot, and even the thing being used as a point of reference and every single instance will differ so it's not actually possible to set a baseline "it needs to be THIS big/ratio to work" because even then it might not be the case or wrong, it's pretty case dependent. And of course, there's also the fact that you always want to actually scale the most prominent thing of reference on that plane, of course scaling off say, a finger nail, won't be as precise as just doing the hand or arm, that can be because of the aforementioned issues, or even just the artist drawing it less detailed to a lower LoD, or anything inbetween.
This feels like a common sense thing and if someone does it poorly, a CGM can just go "btw you should do this, you inflated it, get back to the calc mines and try again", like we do with every incongruity or iffy in a calc.
A suggestion or just a heads up on the px scaling page would be aight tho, nothing wrong with explaining that this can be a thing, and to be careful as to not overinflate values if it doesn't add up or alternatives exist.

This is my opinion, don't need a rule, can't make a rule, but a suggestion or heads up is aight if we really need to take action.

 

[Agnaa]

I agree with Chariot.

First calc linked here is already fixed, so it no longer makes a good example.

And the third calc uses the eye since it's the only reasonable-sized thing in the frame, so it's also not a good example.

Second calc should be changed, since something like thigh length could be used instead.

 

[KLOL506]

Was gonna say something of my own but Chariot pretty much covered all my topics with more detail than I could ever fathom, he just comes prepared for everything.

 

[Vzearr]

Asked three mods, got perm, can show if asked.
I ran into that problem just the other day ironically.

But there's an easy solution. Just use don't do it.
It's only an issue due to ratios/resolution (assuming it isn't art inconsistency because that can in fact lead to off values, simply because the artist didn't draw it perfectly proportionately accurate, which happens a lot) or angles and depth which is why we sometimes use trigonomics to get true sizes and distances if things slanted, at angles, skewed, etc, which obviously isn't the pixel scalings fault, but rather just how depth and angles work, which can be accounted for.

Say we have a 1000px image, and the software, having to use and count in pixels, says "yeah this 50cm object is like 2px", you're forced to then treat every px, using that thing as a point of reference, as being equal to 25cm. And sometimes that might not be the case and it's obvious, that's just a limitation via art, software, and so forth.
But in that scenario just figure out a different path, it's easy to solve and avoid, and the margin of error decreases very, very, very, quickly the more higher resolution/higher overall px count an image is in response to the proportionately small thing being scaled (In most cases, in some odd cases for like live action, camera lenses and filters might skew things ever so slightly but that's something else entirely).
And then there's cases where where a small object, being small in pixels, is proportionate, accurate and scales properly in respect to the other aspects of the image, which is also quite common so while this can be an issue, it isn't always an going to be one.

Don't think we really need to set solid ground rule, mostly because it's completely impossible and it depends entirely on the image, panel, screenshot, and even the thing being used as a point of reference and every single instance will differ so it's not actually possible to set a baseline "it needs to be THIS big/ratio to work" because even then it might not be the case or wrong, it's pretty case dependent. And of course, there's also the fact that you always want to actually scale the most prominent thing of reference on that plane, of course scaling off say, a finger nail, won't be as precise as just doing the hand or arm, that can be because of the aforementioned issues, or even just the artist drawing it less detailed to a lower LoD, or anything inbetween.
This feels like a common sense thing and if someone does it poorly, a CGM can just go "btw you should do this, you inflated it, get back to the calc mines and try again", like we do with every incongruity or iffy in a calc.
A suggestion or just a heads up on the px scaling page would be aight tho, nothing wrong with explaining that this can be a thing, and to be careful as to not overinflate values if it doesn't add up or alternatives exist.

This is my opinion, don't need a rule, can't make a rule, but a suggestion or heads up is aight if we really need to take action.

I believe a rule is needed, it's just a hassle for cgms to keep correcting it. A heads up would come in the form of a rule imo.

 

[Flashlight237]

Pixel measurements tend to be a little finicky. As someone who uses pixel lines, for example, something I've noticed is the lines tend to measure shorter than the actual height of an object. This tends to start being statistically noticeable when objects are 20 pixels or shorter, which is why I use the rectangular selection tool for such small sizes.

That's only from personal experience. I'd say a good rule of thumb is the bigger the reference object is officially speaking, the more accurate you're going to have a calc. In this calc, the tallest reference object usable is a fairly tall man: https://vsbattles.fandom.com/wiki/User_blog:Flashlight237/Re-Calc:_Maple_Uses_Machine_God

Whereas here, the biggest usable reference object with an officially set size is the track: https://vsbattles.fandom.com/wiki/U..._Survives_a_More_Direct_Antler_Bomb_Explosion

I say this because bigger objects are more visible, thus you'll see them more cleanly.

 

[Vzearr]

Bump.

 

[DontTalkDT]

Finding something of a decent size to scale is certainly one of countless considerations that go into selecting measuring sticks.
So, while I agree with the premise in principle, I don't think making a rule like "use the biggest option to scale from" is a good idea.