The "What Are You Working On?" Thread

[Modek]

View attachment 516341

Now we need a Car washing machine in Assetto

 

[Kyuubeey]

Still just grayboxing and making the layout. Hardly a track yet. None of that terrain mesh is final, it's more for visualization purposes.

EDIT: I think that's it for tonight.

 

[NightEye87]

How much caveman am I if I don't recognize the track?

 

[Kyuubeey]

How much caveman am I if I don't recognize the track?

It's Sportsland Yamanashi. The only version(s) we have in any sims are quite bad, so I suppose someone has to make a better one. The track itself has been promoting themselves a bit with some nice footage and Alexi Smith (Noriyaro) has made a bunch of videos featuring it; and it's not incredibly big or long, so I figured there is something to work with.

Made the first hairpin area road and terrain a bit more accurate. Soon I'll have to figure out how that area fits in with the pits.

 

[Thrashy]

Onboard around Lake Garnett:

Looking a damn sight better (and a lot less zany!) than this time last year, that's for sure...

 

[Kyuubeey]

Made some substantial improvements to the main straight, first and last corners. Aerial image I used didn't have a clear view + distortion so it ended up being incredibly off. Now the parts at least somewhat fit in.

 

[Amir Vodokotlic]

View attachment 516427

Made some substantial improvements to the main straight, first and last corners. Aerial image I used didn't have a clear view + distortion so it ended up being incredibly off. Now the parts at least somewhat fit in.

Are you by any chance building this by eye?
Just thinking since Japan has no lidar data or elevation whats so ever.

You get get atleast elevation data from google maps 3d. Should be able to import 3d model into blender granted it won't be scaled but thats easily can be done with reference from know object around track.

Tutorial here.

 

[Kyuubeey]

Are you by any chance building this by eye?
Just thinking since Japan has no lidar data or elevation whats so ever.

You get get atleast elevation data from google maps 3d. Should be able to import 3d model into blender granted it won't be scaled but thats easily can be done with reference from know object around track.

Tutorial here.

By eye, later I'll (try to) photomatch it a bit more properly once things are close enough.

I got shown some Google maps elevation data, but it was pretty horrible, so it's quicker to just do it by eye. The track isn't massive and you can see most of it at once from the tower or from a drone, so I think it's doable, as long as you lower your expectations a bit.

Elevation on the 1st corner was entirely off, so I had to adjust most of that area. Soon it's probably about time to start building the extended course, it should help me place the 1st corner as well.

 

[Amir Vodokotlic]

By eye, later I'll (try to) photomatch it a bit more properly once things are close enough.

I got shown some Google maps elevation data, but it was pretty horrible, so it's quicker to just do it by eye. The track isn't massive and you can see most of it at once from the tower or from a drone, so I think it's doable, as long as you lower your expectations a bit.
Elevation on the 1st corner was entirely off, so I had to adjust most of that area. Soon it's probably about time to start building the extended course, it should help me place the 1st corner as well.

Good luck.
Yeah you're right.
Google data looks to be 30 meter resolution just had check.
USA regions are lot more accurate these days.

 

[Johnr777]

Are you by any chance building this by eye?
Just thinking since Japan has no lidar data or elevation whats so ever.

You get get atleast elevation data from google maps 3d. Should be able to import 3d model into blender granted it won't be scaled but thats easily can be done with reference from know object around track.

Tutorial here.

Actual google 3D is done through photogrammetry and is pretty accurate. I did some data grabbing of Trois Rivières in Canada and it matched the lidar perfectly.
Unfortunately none of that seems to be available in Japan

 

[Amir Vodokotlic]

Actual google 3D is done through photogrammetry and is pretty accurate. I did some data grabbing of Trois Rivières in Canada and it matched the lidar perfectly.
Unfortunately none of that seems to be available in Japan

Yeah seems some places with new google scans are really accurate.

Didn't the guy that built Kyalami track for AC use google 3d data along side lidar to build it.



wonder if taking drone footage and running it with meshroom could be done.

 

[Kyuubeey]

First version of the extended course. Probably way off after the first lefthander, but anyway, first version. Some really useful drone footage and car footage from low to the ground used.

 

[Thrashy]

wonder if taking drone footage and running it with meshroom could be done.

I've tried this a few times, with very mixed results. Of the tools I tried RealityCapture did the best job of reconstructing the drone's flightpath from just the video, but even in the best case it didn't produce useful data for me (in my case I was using drone imagery of the Nashville street circuit to try and reconstruct the repaved portions that weren't in Tennessee's public LIDAR data). Ultimately, in order to get to a decent-quality mesh, you need to have the GPS metadata associated with each image or else the algorithms seems liable to misconstrue features in the imagery and start chasing wild hares -- espescially in the context of a road, where the paving is very same-y and often the track verge doesn't have much in the way of landmarks, either. A related issue with drone shots sourced from Youtube, in particular, is that the compression eats mcuh of of the fine detail in the road surface, which means the algorithms don't have many reference points left to reconstruct the road itself.

You might get lucky and get a few decent sections, though, and that may be all you really need in the circumstances.

ETA: This is not to say that you can't get good results with something like this process -- ToddK has done great stuff with a GPS-enabled action cam, for example -- but that the pose estimation algorithms used by photogrammetry software are just too error-prone without location data to keep them sane.

 

[Mitch9]

Actual google 3D is done through photogrammetry and is pretty accurate. I did some data grabbing of Trois Rivières in Canada and it matched the lidar perfectly.
Unfortunately none of that seems to be available in Japan

The only japanese track with photogrametry available is fuji... just saying

 

[Bozian]

Onboard around Lake Garnett:

Looking a damn sight better (and a lot less zany!) than this time last year, that's for sure...

That looks absolutely incredible! Very nice work!

 

[Trava]

Almost the entire Japan has 5m aerial LIDAR available. It's only ground level though and too low res for any serious project.

 

[Kyuubeey]

Almost the entire Japan has 5m aerial LIDAR available. It's only ground level though and too low res for any serious project.

Bro What Are You Talking About Its Fine ????

 

[Pessio]

Here we go again.
This lil'menace needed some love, was an ugly mess. Let's call it a "remaster"

 

[Kyuubeey]

Made some big changes to the extended course area and lengthened the front straight and downhill straight before the last righthander quite a bit. It should roughly line up to measurements now but I don't think anything's really correct enough yet. Last corner radius is suspect.

 

[Thrashy]

Finally getting around to replacing the placeholder road textures. The park around Lake Garnett was a Depression-era public works program, completed in 1936, and the heyday of the races there would have been right around the tail end of the life of the original asphalt. The historic aerial images show a pretty bright road surface rather than darker new asphalt, so I suspect that the original, aged surface was likely in place for most of the road surface. I've tried to reflect a fairly degraded asphalt condition in the textures.

I've also managed to scrape together enough vintage family photos and Sears catalog models to assemble a reasonably convincing early-60's American crowd, and with a few tips from @Fat-Alfie and the info that @AlleyViper provided in the Bremgarten track thread, I've got custom normal maps going for them as well.

Contemporary newspaper estimates claim 75,000 people were in attendance in 1963, but that seems pretty suspect given than the same year's SCCA meet at Bridgehampton, in a much more populated part of the country and featuring even more big name drivers, only drew about 10,000. In any case, in the interest of performance I'm starting at a few thousand spectators. Still looks plenty packed from the driver's seat.

Plenty of material refinements and performance optimizations happening, too. I think I may not be too far off a preview release.