Heh yeah that most deep learning stuff degrades massively if you switch from one camera to another model is telling. But its still important for most things that arent trivial dl.
Rectification is not always the best choice, its a simple one for data processing in some cases, but often limiting. Keeping to a simple model is much more important.
Calibrate absolute scale to a margin where people can no longer tell most likely.
Its a huge difference between that and you can do stereo at half a kilometer.
I guess you havent actually tried to do this, or you would emphasis the crucial important of, you still need a good initial guess, you still need to know the sync, and you still need data from the right environment with the right motions, and in the end, it will be worse.
Thank you for trying, I thought about it myself, after all someone has to teach people so I dont get another, yeah we collected the dataset with a calibrated camera rig and here it is. What do you mean what was the calibration?
You cant compare a real technical field with hundreds of years of real world use to some social science nonsense. Linguists have never made useful or testable predictions of any kind. Geometric computer vision has been refining camera calibration to the point where we use stereo to find things billions of light years away.
Thank you for keeping us old school computer vision guys in buisness. If deep learning folks actually used good calibration, and didnt constantly argue eh an offset of a few pixels does not matter, they might actually be competetive with classic methods by now. Well competetive accuracy wise, but not performance wise at least. Most VO systems localize a camera before the images have been uploaded from ram to the gpu, and deep learning will have a hard time getting around that kind of performance limit.
Yeah, you can do live simultaneous calibrations and reconstructions and localizations. Its a difficult problem, and a slightly different one for each phone, but there are plenty of variants which work for single phone models. The imu is key to making it possible though.
What makes it harder is that older phones have insufficient processing power, whereas newer phones have too many cameras. The phones also do post processing in undocumented ways that are impossible to disable. There is also rolling shutter. This in turn can be simplified down again by just subsampling the images to 480p and estimating camera parameters per image.
Richer models that fit better are often a trap, in particular for beginners. Use the simplest model you can get away with. Unless you know what you are doing, most people are better of with a simpler model as it will be more robust to observability issues. If you dont know what those are, use a simpler model.
Uncertainty propagation is very difficult to use for vision, and largely just modelling errors in vision as the error distributions, e.g. for anything observed or reconstructed from images are either subpixel accurate, or too non linear.
Opencv is fine for basic models like the most popular one. This one is also sufficient in most cases, I have used it for subpixel 400m stereo with a 40cm base, visual odometry, high end 3d reconstruction etc, it also has fully functional variants for wide and fisheye. You just have to select the right model, calibrate without observability issues, verify the calibration, understand the camera, etc. The better tools help a little bit, but they do not eliminate the need to understand what you are doing in the slightest. The better tools are easier to use, but generally not more accurate. The spline field variants are more accurate if you know what you are doing and the right conditions apply, but much less robust if you dont.
Opencv calibration is hard to get right if you dont know what you are doing, and at first glance it will look right despite being irrecoverably bad. I had to edit the docs and tutorials because the examples they had were of FAILED calibrations. Visualizing the resulting distortion field is critical to understand if the calibration succeeded, and that s what the better tools provide. That said, if you dont know you are looking at, they likely dont help, and the golden rule is if someone calibrated something and said it was easy, they didnt. If someone has a pdh in camera calibration said they calibratied them but that they havent used them for sfm but they plan to soon, odds are 20% its right.
A better way to tell if your password is secure is to
download a gpu dictionary password cracker, and add in a couple of the big leaked password dictionaries, get at least the big ones so you have a few billion passwords and run it for a few hours. It will crack millions of clever passwords like just like that, and the only result is making someone like me wonder how can 7 people come up with the exact same seemingly unintelligible string that long.
This is correct, its not needed for good performance, and most things websites require cookies for dont require the disclaimer, with two caveats. First its pretty clear from the law what requires it, but what if you are wrong, what if corp requires legal to sign off on the engineers statement that it is legal, its not like you gain much from not having the disclaimer, and second, most freemium analytics (e.g. googles) and all ad related services do require it.
FCC the regulator agency charged with regulating telecom companies once again fails to regulate telecom companies. Despite a trivial and permanent remedy beeing available, e.g. fine the telecom companies in significant excess of what they made from providing the robocalling companies with the services required to make the robocalls, they instead place a ineffectual and unenforceable fine on a couple of soon to be chapter 11 LLC. Yay for telecom lobbying....
These are the same assholes that let you use a email containing a + in the email. e.g. name+shittyuplay@gmail.com, and
Automatically log you in after registration, so you can register and buy games.
Then dont allow you to login or use the support services later on because the email contains a +. So I get their emails, including eventually one about the cancellation, but still havent been able to login or use the support for my account. And if I post on the forums I just get the shitty advice that I need to login to support services or request a password reset because neither of those allow + in the email either. I also received what looks like a ban for using curse words in my support request at one point, and yes I did use <name>+shittyuplay so well thats true, but if thats the damned problem dont let me register it in the first place.
I dont care anymore since the game I registered was just some shitty thing that came with a graphics card, but I am astounded at how incompetent they can be while still trying to become more popular than steam. Steam only needed to provide better service than pirates, but Uplay needed better than steam, so well they are doomed and good riddance.
I think its more along the lines of things are prototyped without types, the proof of concept mostly works, then people realize that types are needed to reduce bugs, improve speed, and create a maintainable base as the project grows. What people should do is rewrite from scratch using the prototype as a vague guide, but instead people try to desperately fix the prototype incrementally. Its not the types that makes things slow, its the size of the codebase, and the lack of expectations on quality and performance.
Arguably C is losely typed rather than strictly typed. int for example is some integer type, dont worry about which one, and people keep reinventing this too. Sometimes you want stricter, sometimes more loose, but most languages have a sweetspot where they provide a decent syntax and anything outside it is crap.
The language which solves this will allow people to select the level of type abstraction on a per task basis, while retaining full performance. It will be nice when we get this, but no language is even close.
A single number to represent a rotation in 2D, vs a single number to represent rotations in 4D for which 2D, and 3D are special cases.
Compared to 3 numbers and a series of operations which happen to rotate a 3D vector.
Regardless I think euler angles should not be taught except as a dont do this, its a bad idea. Mostly because when used in practice, in order to interpret them you must convert them to quaternion to do so, which is very telling. Gimbal lock is a secondary problem to that.
The math.h cos from the article counterexample is pretty trivial, so can you give an example? I doubt linear interpolation between the 8 values is enough, but cubic probably would be. This would also be close to e.g. 8 third order taylor approximations of the function, and a comparison would be interesting.
The equivalent of polar coordinates in 3d are quaternions. Which do not have gimbal lock, the problem only occurs for euler angles(which should never ever be used for anything ever).
A good idea is to not to compute the values of cos from 0-2pi, but further reduce the range, using
cos(a) = cos(-a), and cos(2a) = 1-2cos(a), or cos(a+pi/4) =...
So we really only ever need to be able to compute cos in the range 0-pi/4.
Then for further accuracy we can do the taylor expansion around pi/8. (or other approximations)
finally the number of terms for a fixed accuracy varies with the distance from pi/8,
Yes, and had they continued to insist, showed their guns, and refused to leave he would have gotten away with getting a weapon and shooting them. But most homeowners are both unwilling and unable to take down a couple of armed thugs, assuming he even had a weapon at home that wasnt a fantasy sword replica.
Had he escalated and then drawn a weapon, odds are they would have shot him and may well have gotten away with it, claiming they were just asking questions and he immediately drew a weapon and tried to shoot them.
Remember that when some rednecks hunted down and killed a kid a while back for running while black(Ahmaud Arbery) the police initially wanted to drop the prosecution, and its quite possible that without the national attention of the Floyed case, and the murderers being dumb enough videotape themseles, they may well have gotten away with it. The simple fact that it was a discussion of if they would get convicted at all is just as telling. And police will generally back security companies more than rednecks in court.
Its called fait accompli. If I go to someone who stole my bike, beat them and take the bike back, yeah I do get to keep the loot, even if they prosecute me, which is unlikely, and even if they succeed in prosecuting me which is also unlikely, I still get to keep the loot. Its not the way the legal system or the law is supposed to apply, but in reality the idea or intent of law is irrelevant, its its practical application that matters. It is why I explicitly say they get away with it because they mostly do it to criminals.
No one in their sane mind think the law is intended to allow the threats they made etc, or that they had the rights to do what they did, but they did it anyways, and as the story shows, they got away with it. Think about that again, they did this to some rich white geek with a platform and they got away with it. What do you think they would get away with against a known criminal...
Those protections are not the only reason the police are comfortable doing things like that, those protections are there, because the justice system could not get police to stop doing that regardless, and in extraordinarily rare cases its justified.
Nothing that isn't stories by friends working for securitas (the main company for which pinkerton is a brand), or anecdotes. I am certain I could find a story like that just by googling for it, but real statistics are what are interesting, and I cant be bothered to research it. So let me instead ask, what evidence do you have that they are better than the police? because that is what you are arguing. In general you can become a security guard in a tiny fraction of the time it takes to become a police, and they have much lower requirements compared to becoming a police. Oh and by the way, most of them believe themselves to have the same powers and protections as police, that they do not does not matter much. The real shitheads get pruned pretty quickly, but the middling level assholes instead sometimes get shifted into one of the offbrands.