Friends! We need your help with sharing (retweeting, posting the link to) this video, so other open-source photoscanning folks can find it! Can you do it for us plz? :)
We will upload this updated version (that you see here) for the course owners as well. An interesting side effect of adapting the course videos to RU-vid is that the videos get back, updated, reworked and extended, like this one. And hell, anyway Meshroom open-source workflow *MUST* be a freely available video :)
Photoscanning tutorial timeline: 00:00:00 - Start 00:00:54 - Capturing the photos 00:01:57 - Developing RAW photos in Dark Table 00:10:22 - Reconstructing 3D scan in Meshroom 00:15:19 - Fixing the orientation (SFM Transform) 00:17:58 - Depth map 00:20:33 - Meshing 00:22:30 - Exporting the .obj from Meshroom 00:23:23 - Mesh cleanup in Blender 00:26:10 - Sculpting in Blender (mesh denoising) 00:30:18 - Texturing in Meshroom 00:37:25 - Result of the Part 1 - Highpoly 00:38:31 - Part 2 00:40:00 - Mesh optimization - Lowpoly 00:44:51 - Baking Displacement map in Blender 00:54:42 - Normal map 00:56:33 - Diffuse map 00:59:31 - Setting up the material 01:00:23 - Cycles micropolygon displacement 01:02:24 - Photoscanning tutorial result
Thank you so much! Now that Reality Capture is free, the utility of this workflow (with Meshroom) can be compared against the RC workflow, but anyway it's good to have alternative options I guess!
Oh wow, that's impressive! I know that this workflow can get quite involved and even tedious and it could take some time to get familiar with it, so I'm glad this video helped somewhat :)
Really? That is so pleasant to hear, I've always wanted to do something that involves narratives and storytelling, so it's so kind of you to mention that aspect!
Thank you for this video. I've watched a few and it's very clean that you understand why you are doing every step and you communicate that rather than just presenting a half understood flow like so many others. I'm looking forward to reprocessing my last scan using what I've just learnt.
Thanks this is gold :D. The Tiffs I exported didn't have Metadata at first BUT if you go to export tab and hit ≡ preset button > preferences.. > you can checkbox exif data and any other metadata as well! Just posting in case someone is having trouble with camera data not exporting with tiffs
I had that problem and I was using exiftools to copy the exif data from the original images. Lol. I thought it was a bit stupid to do it that way. Thank you.
Hey Gleb .....this is awesome, i am into photogrametry for some time now. I am very much inspired by your workflow and results. When trying to learn about photogrammetry there is almost no tutorials out there for taking proper pictures. So man, it would be so useful if you make a quick tutorial or tips on taking photos for photogrammetry.!
Vijay, we have covered it extensively in the course, but to start with, watch this tutorial: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-JLdxBtECGuc.html
First of all, thank you so much for making this amazing tutorial just a note for anyone having a problem using LSCM unwrap, just change it to basic and it will work fine, I don't know why is this happening think its just a bug
LSCM texturing fails in latest versions of Meshroom. Do you know of any fixes for this? Is there a way to combine the textures from the BASIC unwrap method?
So you're going to update this section within the Photogrammetry course as well!? I've just started and I'm already getting updated content...woohoo! You're doing the lord's work, mate! 😍
We thought it's about time to release a sequel (or is it remake?) of our 'free photoscanning' tutorials. 100% free pipelines are tricky to setup, so the usefulness potential is high.
I'm stuck at the Texturing2 node because LSCM doesn't seem to work. I even played with the decimate settings to bring the model down to around 300k faces but LSCM still fails. Basic works no problem but then I'm stuck with 3 EXR texture files instead of one. I've seen posts from others who are having trouble with LSCM and also with 2021.1.0. Any ideas? (Great video btw, really clear and easy to follow)
Hi Gleb, awesome video, again. A pleasure to learn with the free software community. Many thanks for your teachings. Very well explained! Eskerrikasko!
The understanding that someone might need it keeps us doing it, otherwise proof-reading and manually fixing subtitles is so tedious, honestly! :) Hours and hours of work and it's hard to tell if it's really useful.
Holy wow man. Recently picked up a nicer camera as my phone was causing a lot of frustration. There's a lot I didn't realize I could/should be doing to improve my workflow. I really appreciate it! Subbed
@@GlebAlexandrov Your explanation is fine. I got a bit stuck on the new commands that I don't know because I'm not familiar with Darktable and Meshroom. I can re-watch the video, so it's not a problem.
Gleb Congratulations. This video is a monument! Thanks a lot for your tips....I make photogrammetry with 3DF Zephyr but I think to Meshroom now because you! At soon....
No dude Zephyr, reality capture and agisoft all of them are better than meshroom. What takes 2 hrs in this software would take like 6-8 hrs in meshroom and the resulting final mesh would have more problems and holes in them
Thanks for this. I've just finished a lighting rig and turntable system to scan objects. I slowdown was going to be Meshroom nodes, but this will help with understanding expectations.
Hi John, glad to hear that! If Meshroom proves to be TOO complicated, I recommend trying Metashape or RC to see how it works for you, just for the sake of comparison/research.
@@GlebAlexandrov thanks for that! I'm interested to see how the mesh noise looks. I'll share my experience feedback shortly. I'm using a similar workflow as you except with Lightroom. As as you've so rightly pointed out, a project's top challenge is often workflow! Take care and thanks again.
Thank you so much for this video! I've finished to do my first photogrammetry of my sculpture. Well, the sculpture need more sculpting process in Zbrush, but the main forms are very well. Probably I'll bay your course later.
Timofey, I'm glad to hear your first attempt ended up successful! Photoscanning sculptures is so much fun, when I told my son he could get his clay sculptures into computer, he's now asking me to digitize it :)
We thought we must make this alternative workflow available for everyone, even though we pursue a different one (involving Reality Capture and some other software) ourselves. Meshroom and open-source cross-platform apps in general are the future.
Thanks for the video. The Project files aren't available in that link any more. It'd be great if you could share it for us to follow along with the video.
I cant agree with the method of optimisation. It's better to retop it by hand, make it somewhat quad like low poly model, unwrap it with uv seams properly... so u can bake photosacenned textures and make some variations of that asset with substance. It is not that fast but you endup with better asset that not just look good, it also demostrate you skills as an artist.
I agree, manual retopo with the proper uv seams leads to better, more controlled results generally. We try to use auto decimate/simplify/remesh options when the asset allows it.
Hola Gleb, greetings from México! I purchased this course as an early bird and I'm still looking at it, it is so dense and basic at the same time, as photogrammetry is. I'm an archaeologist and I use photogrammetry for my surveys and research. So, I constantly work with very complex meshes and non-symetrical shapes. Usually, the optimization of the HP is very very hard to do. I've tried Zbrush (z-remesher and dynamesh); I've tried Instant Meshes and now I see your workflow in Blender with Decimate Mod. The question is, what about the diferrence between quads and tris? Because sometimes I found myself in the need to retopologize de mesh to cover holes or segments of the mesh that need a hand, and working with quads is a bit more easy. The second question is a bit obvious, but I think that this workflow (in Blender) is more extreme to for the PC with very dense meshes, from 20-50M polygons, because you decimate and then you subdivide and so on. This processes are very harsh for the GPU and CPU and eventually causes crashes with no further road. Obviously you need a very strong computer to do so.
"so dense and basic at the same time" - perfectly describes photogrammetry indeed, easy to learn, hard to master :)) "what about the difference between quads and tris?" - each quad is composed of 2 tris, so that's an abstraction meant to ease modeling a bit, just like you've mentioned. What was the second question, what's the better workflow for working with multi-million polygon meshes, right?
@@GlebAlexandrov yes, Gleb. That's the question. As a person that works with cultural heritage I need to keep the original mesh intact, so with some decimation tools I see a bit of 'smoothness' or changes in topology, that happens when you exaggerate the decimation.
Thank you so much for your tutorial, Gleb! it is very helpful! By the way, I tried to calibrate the color in Texturing image, which is explained at around after 36 minutes of your video, but it doesn't work as you did. I put the path of the 'meshroom_texture' folder in the duplicated 'PrepareDenseScene' node, but I couldn't compute it. Is there something to be considered for my problem? I am using 'Meshroom 2023.1.0'. Thank you again for your tutorial!
Hi Greb. Great tutorial (I watched at it only now), as usual. A question: I have takena video around a statue, turned that into images and used the frames in Meshroom. The point cloud looks OK, but the mesh only reconstructs the environment around and not the statue. If I constraint the meshing with a box I basically get a box as a mesh. Any suggestions?
Helo Gleb, it is very nice from you to share your knowledge this way, it was a very didactic and complete explanation about how to work with these tools. Thank you very much for your kindness. I have a technical question: there exist, under this meshroom program, the posibility to take pics in diferent positions (X,Y,Z) i mean to turn upsidedown an object to get the belly and all the details, or it is just one position to get all the pics? And in the case that you can get all the positions (X,Y,Z) how you can work with them to get the final 3D model? Thank you very much, Best regards Rodrigo.
very informative and covers some advanced workflows. I also use a two-image set approach to creating models, although have avoided sharpening as was concerned it might create false data points, have you compared the results of both sharpened and unsharpened?
I hope you don't mind all the questions but I now have two mesh files (original and augmented) and want a final combined one to proceed with Printing. Can You direct me to a tutorial on how to get the two combined into one? Thanks
Hi Gleb, thank you so much for this in-depth tutorial. I was wondering if you still have the project files somewhere? The Google Drive link in the description looks broken. Thanks!
Hey Gleb thank's a lot for this video now i use meshroom all the time ! I was wondering if there was a way to save the nodes settings on meshroom, to not redo all the changes? Thank's again have a great day
Привет! Как поступить, если в файлах кадров нет данных EXIF? Например, если кадры получены не с камеры, а путем раскадровки видео. Столкнулся с тем, что без EXIF Meshroom не может сообразить, как реконструировать положение камеры, и вместо ожидаемой модельки получается сборище сфер.
That's something that a) I have very little knowledge of and b) I'm extremely interested in learning more about. I also have a feeling this might be the key thing to mitigate some major photoscanning challenges, e.g. using the lower quality input and just rebuilding the badly reconstructed parts without going into the weeds of laborious manual mesh editing. p.s. Maybe you have some links to share? ;)
I have a question. How to adjust the scale of a model? I want to get a 1:1 model,but it always gives me a 1:650 model. (sorry for my poor english,this is google translation)
Yup, the gains aren't significant, at least if the surface has enough textural details. I've had a few cases where it made a difference though - for example when scanning a mostly white uniform wall (16 bit export produced slightly better results).
imgur.com/a/8aTESML - here I made a graphic illustrating the Lightroom settings that give the result similar to Darktable (with the settings we use from this tutorial). Basically, the color profile is set to Neutral, the highlights to +70 or something, the shadows pulled up, and the Tone Curve adjusted in this kind of way.
Do you know if it is possible to use Mushroom on a bunch of digital pictures... pictures without shadows or backgrounds? I have 60 images of a 3d rendering from every angle and I would like to convert it back into a 3d rendering. Thanks
