Steve Lund dives into Photogrammetry with the new Open Source tool Meshroom and takes you through the steps to making your own first succesful 'photoscan'. See also this topic on Blender Artists if you need support.
blockquote>
Learn how to do photogrammetry and create some amazing 3D scans with Meshroom and Blender for free in this tutorial! All you need is a camera and a computer!
And here's the result of this project:
18 Comments
Hi, thanks for the tutorial , but i cannot do so well. This massege appears when i clic in the start button: "Sensor width doesn't exist in the database for image". I delete many images and clic start again but the massage appears again and again whit a diferent image names
David, I´d like to talk to you.. can you send me an email... [email protected]
This massage appears to: "Please add camera model(s) and sensor width(s) in the database". Do you know how can i fix that, i seen the camera model in the code but i dont know how can i open the database for Meshroom. Sorry im not even know where is the database.
Your need to edit the file "cameraSensors", you find it under "..\Meshroom\aliceVision\share\aliceVision\".
Then you enter your Camera Brand, model and the Camera Sensor size.
As an example for smartphone Redmit 4x
"Xiaomi;Xiaomi Redmi 4x;6.4"
The sensor size you must search for it on the internet
Hi MYUQIYU, thank you for answer me, but maybe thats not the problem because i open the file cameraSensors.db and i find the model and the sensor size of the camera that i used in the list that appears in the file. Do you know any other solution? what else can be plaese?again, Thank you for your help
Try to make an entry without the "/ 1100D". On your first image the program shows only "Canon EOS REBEL T3" in big letters.
So you can try with the entry "Canon;Canon EOS REBEL T3;22.2"
My excitement level went from "super awesome" down to flat line when I realised halfway through the process that I need CUDA for it to work :(
I managed to run the modified version of Meshroom, with bypasses
the need for CUDA, as described here -
https://github.com/alicevision/meshroom/wiki/Draft-Meshing
On my Mac Os X, version 10.13.6, High Sierra, but rather indirectly.
That is, I loaded the linux version of Meshroom onto Ubuntu, 18.04,
running on VirtualBox on my Mac. But by setting up a shared directory,
I now have the results on a successful run in my Mac.
Note: with the modifications above Meshroom will produce some warning and
error messages by, ultimately, no failure.
Now, if I can only figure out how to load it into Blender...
PS: used the data from this tutorial (rock images):
https://www.gamefromscratch.com/post/2018/10/18/Creating-3D-Models-From-Photos-Using-Meshroom.aspx
Will there be a Mac version?
I'd love a build for Mac too, but given that this is an open source project I suppose that's up to their user community to provide.
I tried the last Meshroom with Windows x64 and Linux and it stops working at every node.
I tried fixing the status fil (replacing the "ERROR" line with "SUCCESS" and the 0 value) but it still painfull.
Is there ia good fix?
I have been trying to photscan now with different amounts of pictures but it does not seem to work for me.
I only come to "PrepareDenseScene" after that is stops and I can start it but it stops again 5 seconds later.
What can I do?
Does it show any errors? I ended up having to re-install my Nvidia drivers to fix something with Cuda. There's a topic on Blender Artists here that might be helpful:
https://blenderartists.org/t/meshroom-free-photogrammetry/1120286
very good
Great video on this process! Has anyone had the problem of MeshRoom stopping during processing and having to hit "Start" again? I would assume it should just run through completion.
Also, after doing this I've been getting an error in the console window:
[1/11](2/3) FeatureExtraction
ERROR:root:Error on node computation: [WinError 183] Cannot create a file when that file already exists:
And then this stops it too...
Any ideas?
Thanks
harold
I have successfully gotten it to work, but the result comes out unrecognizable. Do you have any tips for getting a better model out of the scans?
Just downloaded, unpacked the files, added 50 photographs - and waited 45 minutes (i7 2600k / NVIDIA 980). Then I got a fine first scan. No issue.
Used a printed letter as a background.