Install the app
How to install the app on iOS
Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Is this useful? - Generate 3d meshes from photos...
- Thread starter JCitron
- Start date
johnwhelan
Well-known member
Generally Trainz has very specific requirements on the poly count side and texture side. I suspect this tool would be a little like sketchup, fine for static models but not for something that has to be rendered in a sixtyth of a second.
Sorry John
Sorry John
Generally Trainz has very specific requirements on the poly count side and texture side. I suspect this tool would be a little like sketchup, fine for static models but not for something that has to be rendered in a sixtyth of a second.
Sorry John
Well its very pricey if you use the highly specced version. but rather then dismissing it out of hand, it might be better to investigate it ,as it may very well save a huge amount of time in learning complex 3d software.
I found this tutorial , would this do the job of making low poly models ? / https://www.youtube.com/watch?v=oMUGxUqAql8
I was looking at this a couple of weeks ago. The program that I looking at was Agisoft Photoscan, which works on a Mac. The article I read suggested that it was considerably easier to use than 3DF Zephyr, although it is not free. Where it might come in useful is to take pictures from which you could later extract dimensions. It might be useful to create detailed meshes from which simpler meshes can be baked. From what I can see in the examples online if does not produce crisp meshes. So it might be limited to producing organic models such as rock faces, dead trees, etc. But it is not good with fine detail either. The example I was looking at was a flowering plant. It did reasonably well with the leaves but missed the thin flowering stocks. You have to take many pictures so the subject has to be static. Might be fun to play with.
There are a number of free programs out there.
https://www.3dbeginners.com/list-of-free-photogrammetry-software/
3D graphics sure hasn't reached its zenith yet!
https://scanbox.xyz/blog/alicevision-opensource-photogrammetry/
Latest reviews would suggest that COLMAP is one of the best. It's also free and works on both Windows and Mac OSX.
https://all3dp.com/1/best-photogrammetry-software/
Cayden
Interesting comments...
It's cool as I said, and the more I thought about it I was concerned about the number of polygons and the number of photos needed to generate a decent model. I can see this useful to convert a static model railroad object into a scalable 3d-model, but more difficult for a real outdoor asset due to the number of photos needed to get the right shape. With other photos, even Google Map views, those images are only what the photographer supplied and not ones we ourselves took. The problem then becomes what goes on in the back, sides, and other parts of the object, or building we're trying to model.
With polygons, the point-clouds will generate a complex mesh that will have to be decimated to something more reasonable. This process can have adverse affects on models - think of wheels for example that get chunked down way too low.
Other thoughts...?
It's cool as I said, and the more I thought about it I was concerned about the number of polygons and the number of photos needed to generate a decent model. I can see this useful to convert a static model railroad object into a scalable 3d-model, but more difficult for a real outdoor asset due to the number of photos needed to get the right shape. With other photos, even Google Map views, those images are only what the photographer supplied and not ones we ourselves took. The problem then becomes what goes on in the back, sides, and other parts of the object, or building we're trying to model.
With polygons, the point-clouds will generate a complex mesh that will have to be decimated to something more reasonable. This process can have adverse affects on models - think of wheels for example that get chunked down way too low.
Other thoughts...?
I can see this useful to convert a static model railroad object into a scalable 3d-model, but more difficult for a real outdoor asset due to the number of photos needed to get the right shape.
They take hi-res movies using drones and clip stills from the movies to get the images needed for the modelling.
Dinorius_Redundicus
kuid 68213
Getting good photos of all sides has been the bane of my 3D modelling life. It’s good to know that it’s no less a problem for these software automatons.
I came across the notes I had made a few weeks ago on photogrammetry. The webpage I was looking at was:
https://www.imore.com/get-started-photogrammetry-macos
As you can see the process reproduced the plant leaves but not the fine detail of the flowering stalks.
As others have pointed out one needs a lot of pictures. On a Mac there's an app for that. ffmpeg can extract frames from a movie.
https://www.imore.com/how-extract-images-frame-frame-using-ffmpeg-macos
Cayden
https://www.imore.com/get-started-photogrammetry-macos
As you can see the process reproduced the plant leaves but not the fine detail of the flowering stalks.
As others have pointed out one needs a lot of pictures. On a Mac there's an app for that. ffmpeg can extract frames from a movie.
https://www.imore.com/how-extract-images-frame-frame-using-ffmpeg-macos
Cayden
I came across the notes I had made a few weeks ago on photogrammetry. The webpage I was looking at was:
https://www.imore.com/get-started-photogrammetry-macos
As you can see the process reproduced the plant leaves but not the fine detail of the flowering stalks.
As others have pointed out one needs a lot of pictures. On a Mac there's an app for that. ffmpeg can extract frames from a movie.
https://www.imore.com/how-extract-images-frame-frame-using-ffmpeg-macos
Cayden
I've seen ffmpeg it's a nice program. A friend of mine needed to capture some images for a brochure she was putting together. She had a video of the product but no still photos and other methods didn't work because the mpeg files split the image into A/B frames and sections. Using this program, she grabbed the product shots and they looked really great in print.
One of the issues with capturing photos and producing a point-cloud for the meshes is the resolution of the original imagery. If the photo or video is very high resolution, then the object data will be more refined. This is why the flower stalks on the hosta plant disappeared. The resolution was too low to capture that.
Probably not. Autodesk has a competitor program I've played around with. It's results were reasonable for making a generic 3D model of the item (Excavator in my case) but would not work well for any gaming situation without a lot of cleanup/work. It also struggled with holes & transparency.
I think the technology is very promising, but not really to a use-able state yet. We've had high-end lase scanners for quite a while now that can return much more refined results; but even that isn't going to be good for gaming or engineering (why we have them, and not use them at work)
peter
I think the technology is very promising, but not really to a use-able state yet. We've had high-end lase scanners for quite a while now that can return much more refined results; but even that isn't going to be good for gaming or engineering (why we have them, and not use them at work)
peter
This looks really interesting John. I see they have a free and limited program for personal use.
Does it have the necessary capabilities or program format for uploading to a Trainz kuid though?
If so, it may be suitable for static and background models.
Cheers,
Roy