Your Personal Micro Blog(s)

That's an even worse rabbit hole. What I was mainly referring to, would be the theoretical possibility of replicating a real car to [minute fractions of an inch] accuracy in a video game mod, the way some developers do. Trying to get it to drive & deform exactly right, well, that's something I'd want to do, but I have to admit I've no clue as to how. That level of number crunchery doesn't seem like it'd be easy to come by on civilian-level resources, so you could make a close approximation of a car's handling & deformation fairly easily, but dead-on would be almost impossible.

 

One of the first things I was taught on my Uni course, is that games are mostly smoke and mirrors. Which I think is something that most people intuitively understand, even if they haven't thought about it.

Fast approximations are generally better than calculating things in full, since it often isn't possible to calculate something in full in real-time. So games like Assetto Corsa do a really great job of simulating vehicles, but under the hood, it simulates far less than the real world, but it simulates enough to feel legitimate (or as legit as simulations currently go). BeamNG simulates a lot more than Assetto Corsa does, however the result doesn't quite compare in simulation feel, and it still is a very simplified model for calculating things compared to the real world. Perhaps that makes Assetto Corsa a much better optimised simulation for driving feel? Of course the BeamNG approach offers numerous other advantages, and isn't finished yet.

Industry does use crash test software, its a lot more advanced than BeamNG. My best guess is that it has a better understanding of materials, and calculates results based upon its understanding of how strong a material is, and how it is sculpted and formed. But ofc BeamNG is real-time, car manufacturers can wait an hour if it means more accurate results.

http://www.3dexcite.com/en/work/references/advanced-study-of-crash-test-simulations

 

Humans in general like impressions better than facts, so it is no wonder that even within games impressions is what counts, not what really is.

Then it comes to what you want to use game for, doing online racing, AC is incredible experience that will set your senses rocking, but try do a backflip with a car and it kinda lands flat

In BeamNG, level design is area where you see largest part of smoke and mirrors, car graphics do contain some too, but this is different philosophy in BeamNG, they try to make driving as much as possible simulating real things where it counts. Sure tires are not quite there yet, but they have got lot of things developed to very best in gaming industry.

Latest development of FFB is giving that impression many have been lacking, even physics have been there, now FFB is really backing up physics.

FFB can give you lot of smoke and mirrors that make you believe things that are not there, it is tool for magicians

Level design is interesting area as you need to thing likely viewing angles and how to make things look believable when there really is no way to put laser scanned everything on a map, things would lag horrible if you would do that.

Things you have to consider is for example how to limit viewable distance, so that it does not look like it is being limited, while still being able to put lower LOD or cut off rendering completely. There are all kinds of considerations one has to make from limits of software, hardware and skill.

Fun factor has to be there too, but it really needs to be thought with magician eyes, magician has to control what users will see, where users will focus their attention and use that to create illusion.

Learning art of magic is very good trait for good level designer as it teaches some things about how humans work and how level designer can use that to make better illusions.


I think that maybe in a future, we can have software that produces usable meshes from 3D scanning of whole car, but it needs some AI to know which vertexes can be considered noise, which are details, etc. My undrestanding is that at current level human labor is needed to fix the shape so that it is usable as there is so much optimization you can do to mesh or alternatively you can waste polygons for nothing to be gained and understanding of how humans experience what they see, how to trick the eye, that is key to best looks within limits of current hardware.


On the other news, I have had horrible headache, had to resort to painkillers and sleep, so day is a complete waste. Kinda wanted to go to forest and try to grab few shots to attempt to learn how to make a texture from such, but I doubt that will happen as each step feels like hammer hitting something behind the eye.

 

The problem with 3d scanning is that it gives you something that looks like this, which is an absolute mess:


A 3D artist would create a model that looks like this:


So its less a matter of cleaning up the model from 3d scanning, more a case of rebuilding it from the ground up. It should be possible to build an algorithm that takes the 3d scan data and builds a "clean" model for you, but computers lack the visual intelligence of humans, so it wouldn't probably reach anything as nice as the above Koenigsegg.

 

From here, open PDF from link that says Power Crust:
https://stackoverflow.com/questions...or-surface-reconstruction-from-3d-point-cloud

Tensor is about image augmentation, what little I read, but surely there will be 3d point cloud to surface accelerated computing at some point, if not already possible, because robot vision from my understanding is combinations of image augmentation and ray casting (which can produce 3d point clouds).

BeamNG is right there then, simulating computer vision sensors, according to research page.

So at some point, we will see 3D scanning to surfaces in realtime as that is kinda needed for robot vision and that is then what can be used to scan vehicle and put it to mesh too, but from there it is still bit further to get mesh detailed and optimized, but that is coming.

 

Spotted in a Washington Post restaurant review. Apparently we're supposed to think less of a restaurant because it uses straws that a) don't turn to mush in your mouth and b) are sanitary because honestly, how are you supposed to properly clean the inside of a metal straw. Also, "woke?" Is this 2016? I don't usually complain about the Washington Post, but this is just cringe in its purest form.

If it's such a big issue, just make plastic straws recyclable! Actually they are recyclable, but most recycling companies won't take them due to their shape.

 

How do you clean a metal straw? Put it in boiling water.

 

There have been talks of straw bans in the US. I have been taking straws each time I visit a restaurant and am actively hoarding them for the straw-pocalypse

 

ive never seen a straw that isn't plastic.

Metal straw? wtf

 

I would be pretty annoyed if they banned straws, my mustache is fairly long and if I don't use a straw my drink gets on it

 

That's called saving some for later

 

Just figure out a way to recycle them! We didn't ban plastic bottle caps for not being recyclable, after all - in 2010, recycling companies began accepting them.

 

Ironic, because I also have a mustache (not fully grown, but noticeable at the very least).

 

See, that's what I'm worried about... see, in my opinion, BeamNG belongs to a class of hardcore simulators for which a good guess, or even a close approximation of the real thing, just doesn't fit. The problem is that other games actually have an easier time of getting "close enough to feel like the real car" because they just decided long ago that they don't care about damage or what happens to chassis rigidity when you add or remove various parts, dump in a much more powerful engine, etc. They probably also have access to manufacturer data on structural rigidity, which some guy sitting at home dreaming of the day he has the car and skills to make mods wouldn't have. But with BeamNG, suspension action, chassis twist, and body/chassis damage are all handled by the same physics engine, with everything else layered on top, thus both deformation and driving characteristics have to be right at the same time and have the potential to interact with each other., as changes made to reflect the real car's crash deformation characteristics could subtly affect handling. And since there is no real way that I know of to convert from known strength/rigidity data to jbeam values or even any reliable way to make one, about the only way to sync a real-car mod with the actual real-life car would be to watch a bunch of crash-test videos, do the same thing to your mod in the game, and if it doesn't look the same then change stuff until it does. That's what I'm wondering - is there any way to break out of that trap and craft a truly-accurate real-car mod?

As for laser-scanning environments, the only way I'd thought of to deal with the inherent lag would be a priority system to determine which sections can be laser data (I assume it works better for roads than cars, since with roads the surface imperfections are part of the fun) and which ones have to be done more simply:

Priority 1: Freeways and mountain-pass roads selected for digital preservation under what I'm calling the "Great Driving Roads Project" (or the "stop with the JDM spam already project"), along with roads that will be used to craft racetracks in the mold of Gran Turismo or Forza's city circuits.
Priority 2: Major paved roads
Priority 3: Other paved roads & parking lots
Priority 4: Unpaved roads (get a low priority due to the obscene amount of cleanup, janky half-solutions, and just plain guessing that would be required to make them drive correctly esp. in games that don't simulate terrain deformation, which is most of them. Also BeamNG, to the best of my knowledge, still does not properly simulate the physics of loose surfaces yet.)
Priority 5: Non-road, non-roadbed surfaces which could in theory be driven on by someone with a 4x4 or dirt bike. These are pretty much pointless as no one would likely notice if they weren't 100% accurate anyway, but are here as an "eventually" in case someone wants to set up a rally-raid route in game.

For buildings and other collidable non-road objects, the priority is determined based on how likely they are to be hit and whether the surface shape really has enough nuance to justify using laser data. For most buildings, as an example, laser date would really be useful only for getting the position right, as details of siding type and whatnot will have basically no effect on the appearance of a collision against that building.

 

Did lots of troubleshooting with the less crappy Dell I recently acquired. Lots of fun trying to fix a funky screen/graphics card/driver? Not quite sure. It’s fixed-ish for now though, hopefully I don’t jinx it.

Core 2 Duo:

Can’t
Obtain
Really
Excellentperformance

2

Dang
Undeniably
Ouch
becauseitsreallyhot

 

I had a bit of a rough week, or two weeks.
Hope you don't mind me venting here.
I attended Miatas at Sevenum two weeks ago, it was awesome, camping with 200 other miata owners and seeing almost 600 of them on one day (saturday).
But before I left for Sevenum, my mother and I went to the forest with our two dogs to participate in a walk for Bernese Mountain dogs only.
One of our dogs, Hektor, was 10 years old and not feeling well, short of breath and slow. The other one is 2 years old and was too busy chasing a horse to notice.
While I was in Sevenum on saturday, mom went to the vet and they confirmed that he didn't have much air. I went to @Neidjel for his birthday and got home around midnight and found out that our oldest dog's condition only got worse.
We went to see the vet on Monday and they made an X-ray scan and found out that he had fluid pushing his lungs and heart away which explained why he was so short on breath.
On tuesday, the vet removed fluid and gave him medication to remove the fluid. His condition improved a lot which was nice since my brother and my father came back from their holiday that evening. With this positive spirit, I went to gamescom on Wednesday which was nice but crowded as usual.
Also found out that our skoda fabia doesnt get close to the advertised top speed

In the weekend, after going to the scouts(I'm a scout leader), I noticed that Hektor was doing worse, he didnt eat at all.
On Sunday, we made the decision to end his suffering and put him down on monday evening.
Which we sadly did.
For a Bernese Mountain dog, he got very old but it was still sad to see how much he "lost" in his last weeks.
Our youngest dog, Rakker, is also sad. I hope he gets over it soon.

But the rollercoaster hasn't ended yet.

I got the Blueprints of a MG TD last wednesday so I could make some progress for the Fintray

I hope the remaining days of my holiday (this weekend...) will stay a bit more calm..


/vent

 

Today and yesterday, I noticed that the area on the cover where you put the straw in moved from the center to a edge. Interesting...

Was there any reports about that change? I don't really watch the American news anymore.

 

Thats because different companies use different designs. Nothing new.

 

Yeah, but every cover I've seen before yesterday always had it in the center (no matter what company it was).

It's also kinda awkward, especially that I'm so used to it being in the middle.