The "What Are You Working On?" Thread

That's how the video card renders the geometry, NOT how the mesh is built, actually they're all quads.
Nothing wrong with the reflections here:
View attachment 181100

As for the disappointed comment I don't take it as something bad or negative, I'd take classics any time over modern cars and is understandable that at this point is what it's expected of me.
This is a paid job I'm doing so it's not entirely my choice. Would love to do the Steve McQueen car though :)

On the triangle flow thing, do you set them explicitly at any point, or preview them before they're exported to check they give a nice result?

I still see superior normal flow over 'treated' surfaces, than ones where software has tried to get the triangle orientation correct.
Given the quad mesh is so smooth, it seems risky to leave the software to get the triangles right?!

Hmmmmm

Dave
 
Good to hear it worked out! I use it at the start of every project, then at the end before detailing just to make sure all the proportions look right. Is a very useful tool if you get used to it on your workflow.

The Focus is looking spot on, good luck working on the interior details in real scale! lol :)

Quoting from the FRS thread, but the post is probably much more valuable and on topic here.

Are you talking about the built in Max cam match tool still?

How do you figure out real points and their distances (and angles(!)) to each other in 3D space from photos?

Do you use EXIF data and just lining stuff up by eye much? Do you manage lens distortion anywhere? How do you double check the results are accurate (ie, lens distort profiles are correct?)



Did you ever see "ImageModeler", it came with Max subscription around Max9 iirc, and I have a feeling it was a standalone purchase too. It was really very nice as you could simply get a load of photos and pick out key points in them all, and it'd slowly figure it all out.
Basically the early photogrammetry type technology for cam matching, but you did all the feature detection work haha. It was good though because you could explicitly set all the points (maybe with stickers on the car too)
Then you could just take a few known points where you knew the distances between them (say along a panel gap) and get the scaling right too.
Also it was good because you could load in photos of different cars and different lighting, and still see the same features for aligning/reference (unlike photogrammetry which doesn't like any of this)

I'm not sure how it managed lens distortions though, because the cameras it created were just basic ones, but points seemed to match 'ok' in Max with the basic imagery (note top left image, rear wheel isn't correct, probably lens distortion related...)
Only issue with 'web' images is exif/crop issues. You can't fix distortions reliably unless you have the originals I suppose.

a110.jpg


BUT, it's slow work making spline cages from multiple cameras like this, mostly because the camera setting and background setting are tediously slow to manage (I started a script, but haven't returned to this workflow for over two years now probably)

It's certainly the best way I found to work on a car you have good photo access to, but no real access at all.
I have about 30 images in 4k size of this car from wikipedia (iirc) and they're all matched nicely.




Autodesk rant!
I was seriously annoyed when Autodesk just binned this product, while keeping the old Kinetix (although still good for it's 20 year vintage!) cam matcher in new £££ versions of Max and pretending there is nothing better to add to it or supersede it.

Why Autodesk didn't add this matcher to Max, rather than saying 3D Fly (later some other name like 3D catch, and now *another* name!) superseded this product, which it clearly damn well didn't, because photogrammetry is totally different to an image agnostic cam matcher!

Doing a quick read up to refresh my memory, Image Modeller and Stitcher were bought by Autodesk in 2008, they ran them as sub softwares for a few years, then binned the products.
They released Fly and 123catch etc for free, but it's totally different software tech. ImageModeller was unique, and afaik there is nothing else like it around.

Maybe if they integrated what they bought with their existing tools then great, but in practice, the end result feels to be consumers getting worse tool-sets and consumers being seen as cash cows.

So it's good business for Autodesk and the shareholders, but as a chimp who just has a job to do, make 3D content, their 3D tools really aren't as good as they could be, and the market generally is worse in my view, because they suck up useful tools but fail to offer them themselves.

They did the same with point clouds tools too. They bought that pointcloud sw in 2010/2011 maybe. Sat on it for 3 years before it appeared in Max, maybe as a sub only feature. Now 5 or 6 years later we have clouds in sub-only Max 17, and you need other sub-only tools to create the point clouds to use in it!
I've got more point cloud functionality in my old vintage Max version with plugins, than the latest versions of Max! In a time of staggering point cloud progress, Max is 4 years behind the curve!



Arghhh this thread is making me want to go outside now and do some more photogrammetry based testing :D


Cheers

Dave
 
Last edited:
Uh ok, so I wanted to double check using photogrammetry techniques to detect camera positions and camera distortions, to then use those for a basis for modelling off photos.

Since cars don't work with photogrammetry so well, but their surroundings do, you can leverage the photogrammetry process to get good camera locations and undistorted photos for reference for spline modelling and whatever other reference you want.

This will also work really nicely for car interiors.

This probably won't work well with photos you didn't take yourself in a controlled way. It'll get close, but it won't be absolutely perfect like this.

What you're seeing here is a bunch of cameras that have been aligned in Agisoft Photoscan, with my car fully in shot in each one.
The matching has mainly been done over the surroundings as much as the car itself.
I've exported the point cloud to Max, and the cameras. I exported undistorted photos from Photoscan too.
In Max you're seeing the point cloud, and undistorted photos in the background.
In this case I'm fortunate in having point cloud support, but you could easily generate a rough mesh which gives a rough idea of some key car details (tyres for example) to start roughing out the mesh from and working from there.
overview.JPG

1.gif

2.gif


In the past I've done this and had unreliable results. The key was (obviously in retrospect) the fact the camera focus changes the camera properties a bit. Shooting with a fixed focus has resolved that issue entirely.

I've just placed a point on the rear ariel tip, and tweaked it from the angles of different cameras, and now it stays put from one shot to the next!

I'll write up a bit of a 'how to' with some more challenging imagery, perhaps an interior, and how to start working with it just using the images/cameras and spline caging.

But it will need Agisoft Photoscan... but hey. Still useful I'd hope!




Yes you could just do a full photogrammetry of a car, but in practice this is for getting solid reference materials for a solid spline cage of a car where you have no prints, and can't go dusting the car or whatever else.

Ie, a Ferrari Enzo in a posh car garage, this technique will work to get pin sharp materials for spline caging and subsequent modelling. And it won't need you to do anything except walk around with a camera and a tripod.

Cheers

Dave
 
  • Deleted member 130869

Playsure #230 helmet edited, new suits painted for drivers and crew, gloves to match, etc etc etc. Now to just have the time to put it in game to check if lines line up, and if colors look ok. My original pitcrew was pure stipulation but now I have one measly poop photo to go off... not even as much benefit for the driver suit but it seems like it was mostly bland so it will have minimal advertising. The original carset for Redline GTP actually proved helpful here!

Question: Should I add the WSPC sticker on top of the cockpit, which is unlike the only reference I found, or should I add it anyway, to blend in with the others? I am leaning towards not doing it, as cool as it could look (I took a photo of an actual WSPC sticker I have).
 
I see, no problem then :)
Sorry if I sounded a bit rough with the NOT, I've just read what I've wrote and might sound like I was pissed off or something, and that was NOT the case :)

Back on the image reference subject, just replied to the other thread (Sera), photogammetry, ImageModeler or any similar tool can be useful in certain situations. I'll just stick to the basics for cars, works out for me!

Some more progress on the interior, slow work but getting it done. Close to the budget here (less than 190k triangles for the full car right now), still missing seats and some minor details like seatbelts, badges and pedals, but it's close to being done.

screenshot007 - Copy (2).jpg
screenshot011.jpg


Textures on the gauge pod are temporary as well.

screenshot012 - Copy (2).jpg
 
Since cars don't work with photogrammetry so well, but their surroundings do, you can leverage the photogrammetry process to get good camera locations and undistorted photos for reference for spline modelling and whatever other reference you want.

If a background of some grid were allowed to reflect off the car, do you think the photogrammetry would work better?
 
If a background of some grid were allowed to reflect off the car, do you think the photogrammetry would work better?

It would be far better if the grid were on the car itself (tape, sticker, etc). A reflection will be in different spots at different angles, so it may actually be worse.
 
If a background of some grid were allowed to reflect off the car, do you think the photogrammetry would work better?

With the correct algorithm and masking (probably mask the surfaces that you want to calculate as reflective, rather than diffuse), then you could infer the shape from the reflections.

For example, I've seen that process whereby you scan reflections (projector pattern or laser line) over an object, while the object rotates on a turntable.
That illuminated profile or pattern changing vs time can then be used to determine the shape of the object.

This whole space of capturing reality is just waiting for a big change. It's already progressing quickly, but is due to progress at a stupendous rate in the next few years I think.
With cars now getting laser scanners so they can drive themselves, the new iPhone having stereo cameras, and automated systems for delighting and generating albedo and gloss maps becoming more understood and available means we might be able to capture a car fairly well in a few years time just with our mobile phone and an app!

The old ways still work, but the new ways work too... as with everything you pick the best tools for the job and work from there. Having more tools is a good thing :D

Cheers

Dave
 
Well was tinkering with RIR today to pack up the update and realized I wanted to experiment with reverb settings for the back straight heading under the bridge. I didn't just want the bridge like NJMP but also when you drop down with the banks on both sides. I also wanted a noticeable volume increase as you passed under the bridge. I think I got it how I want it now.

 
Well was tinkering with RIR today to pack up the update and realized I wanted to experiment with reverb settings for the back straight heading under the bridge. I didn't just want the bridge like NJMP but also when you drop down with the banks on both sides. I also wanted a noticeable volume increase as you passed under the bridge. I think I got it how I want it now.

Oh thank God... ))phew((
I was so worried it was going to be something different. :laugh: What you've done is nice and subtle. I like it. I think it works. There are a few tracks around where you pass under a power cable and it sounds like you've been driving in an echo chamber. :cry: I just listened to yours a few times with my eyes closed. The effect is there, but it doesn't sound all "check out my FX!!!" I'm sure you get what I mean. I like it.
 
0sm6v1f.png

so today i messed with the "textures" division and was so happy to see the image in the "right place" (2 minute photoshop texture dont blame me) that i wanted to post this progress here

Also the back has changed now modeling wise (reshaped almost the entire back) and theres a new bumper with the difuser

Btw this is just a test ... and first try ever in the UV thingy ...i know its not supose to wrk this way:D
 
Last edited:

Latest News

Do you prefer licensed hardware?

  • Yes for me it is vital

  • Yes, but only if it's a manufacturer I like

  • Yes, but only if the price is right

  • No, a generic wheel is fine

  • No, I would be ok with a replica


Results are only viewable after voting.
Back
Top