Cartoonish boiler effect -- illusion or computing artefact?

  • Thread starter Thread starter Swamp Thing
  • Start date Start date
  • Tags Tags
    Boiler Computing
AI Thread Summary
The video clip in question shows an engine stopping and reversing, creating a visual effect where the boiler appears to bend, reminiscent of cartoon physics. Observers note that this distortion affects not only the engine but also the surrounding environment, including trees and passengers. The primary discussion revolves around whether this effect is a subjective illusion or a result of video processing, particularly the YouTube video stabilizer. It is suggested that the stabilizer misidentified the engineer as the main subject, leading to distortions in the background while attempting to keep the engineer stable. The stabilizer's automatic adjustments can sometimes produce noticeable warping, which can detract from the viewing experience. While video stabilization is generally beneficial for shaky footage, this instance highlights a misapplication of the technology, raising questions about its effectiveness and the need for manual control over stabilization features. The consensus indicates that while the stabilizer can improve video quality in many cases, it may not be suitable for all scenarios, particularly when it alters the intended visual presentation.
Swamp Thing
Insights Author
Messages
1,031
Reaction score
769
TL;DR Summary
The little engine that shimmied.
But why?
In this clip www.youtube.com/watch?v=ZGtUGuAY63E&t=25s starting from 00:25, the engine comes to a stop and reverses. If you watch the boiler, you see that it appears to bend under its own inertia like the boiler on a cartoon engine might when reversing in a hurry.

If you play that again and watch carefully, you may notice that everything is affected, including trees and the people on board.

My question is, is this a subjective illusion in the viewer's brain, or is it an artefact of the digital photography / video compression / rendering process? If the latter, what is going on?
 
Computer science news on Phys.org
It is a very unreal looking effect. So unreal, that I doubt that it is physical. I notice that the comments on the video blame it on the Youtube video stabilizer.
 
Ah, the stabilizer. I should have read the video comments first -- and it isn't like there are too many to read.
 
Hah! That's adorable and hilarious!

My guess (without cheating by reading comments) is that the video stabilizer is keying on the engineer. He starts off vertical, but was tilted forward by the rapid stop;. The stabilizer is trying to keep his shape as fixed as possible. You can see this if you concentrate only on him.
 
I was half expecting the guy to jump off with a big stick and begin beating the thing, " when I say 'Whoa, I mean Whoa.' "

Video stabilizer - what's the point if it warps the scene. Must be a cheap knockoff.
Youtube stabilizer - it can alter after the fact a video, seems preposterous.
 
256bits said:
Video stabilizer - what's the point if it warps the scene.
The recording was born pre-warped by camera action; that can't be helped.

The purpose of a video stabilizer is to stabilize the subject of the video, at the cost of the background.

In this case, Youtube's AI guessed wrong as to what the subject was.

256bits said:
Youtube stabilizer - it can alter after the fact a video, seems preposterous.
Not all videos are postcard perfect. Many would be useless without stabilization.

Granted, this is a poor application of stabilization. Would have been better to turn it off. AFAIK, it is now 'on' by default. When I uploaded a video last week, I saw that it auto-stabilized it without asking.
 
DaveC426913 said:
The scene is already warped by camera action; that can't be helped.

The purpose of a video stabilizer is to stabilize the subject of the video, at the cost of the background.
Yes, useful for moving objects.
I am wondering why someone in the video comments said Youtube video stabilizer.
Camera stabilizer I can understand.

Certainly this effect can cause headaches for viewers as everything is moving that shouldn't be.
A mis-application of technology perhaps.

A few more seconds, there is one individual with what looks like an expensive machine on a camera.
Can we assume the video is shot with similar equipment?
Then we assume that these guys know how to capture decent video.
Would have been nice to see the raw video instead of the "fixed" up version.
 
256bits said:
I am wondering why someone in the video comments said Youtube video stabilizer.
Camera stabilizer I can understand.
?
Why do you find Youtube stabilizer implausible? This is what it does.
 
Last edited:
  • Like
Likes 256bits
DaveC426913 said:
?
Why do you find Youtube stabilizer implausible? This is what it does.
The Youtube stabilizer produced some noticeable distortions.
So why use it?
 
  • #10
256bits said:
The Youtube stabilizer produced some noticeable distortions.
So why use it?
As when you have an assistant or trainee, overall they make things easier. Every now and then they make things worse, but the overall average is an improvement so you keep them around.
 
  • Like
Likes 256bits
  • #11
256bits said:
The Youtube stabilizer produced some noticeable distortions.
So why use it?
Not long after Fukushima happened there was some recordings released about the environment/internals of the reactor buildings. Those recordings were just terrible: like a hand camera with a fisheye optics in the hand of somebody with a serious neurotic disease.
After stabilization and correction - well, it was still terrible but at least you could watch it without getting a headache:wink:
This Youtube stabilizer thing seems to be quite decent IMO. Some things does improve.
 
  • Like
Likes 256bits
  • #12
That's good responses.

As I mentioned, the camera men look professional. maybe, I would have liked to see the original video, unaltered, just to see what they were fixing.
 
  • #13
256bits said:
... to see what they were fixing.

I'd bet my paycheque no person had a hand in this.

IMO, it's software and it's automatic. It answers all the questions in one swell foop. Consider:
  1. The engineer inadvertently tilts forward "in real life", due to inertia, when the train stops. (observable)
  2. Youtube's post-edit software is designed to treat a recognizable human - especially one in full-view, centre stage - as the primary subject. That's a darned good bet, 99% of the time. (known behavior)
  3. The software assumes (this time incorrectly) that the shift of the subject in the video is due to camera shift - not due to "in real life" movement - and fixes it accordingly, so that the subject is stable, at the expense of the background. (deduced from 1, 2)
  4. Youtube gets it right 99% of the time, so setting the feature to ON by default facilitates 99% of use cases - leaving only a handful wherein the uploader must take an extra action to turn it off. (software best practices)
  5. In this case, the uploader neglected to turn off a default feature that's beneficial in 99% of cases. (deduced)
 
Last edited:
  • Like
Likes Swamp Thing and 256bits
  • #14
Your explanation makes good sense.
Thanks
 
Back
Top