How is 3D-scanning useful for 3D-artists?
Hey everyone,
My core question: Is 3D-scanning a useful tool for 3D-artists? If so, what would be your benefit?
Background of my question: I’m part of a small group of three persons who are developing software for 3D-scanning based on our scientific research. Basically, it let's you assemble a 3D-scanner from off-the-shelf components - such as DSLR cameras and projectors.
One potential use-case of 3D-scanning I have in my mind is to use it for 3D-modelling. E.g. real objects could be 3D-scanned in order to speed up the manual creation / editing work, right?
However, I’m not a CG artist! So I’d like to discuss whether you think such a software would be useful or not in your workflow. Output of the 3D-scan would be a mesh (e.g. in OBJ, STL) together with color / normal / displacement map.
A set of example scans can be seen here (raw result, not post-processed): https://sketchfab.com/3digify
Would this be a quality you could work with? One problem could maybe be the messy topology? Could this be fixed? I’m really interested in your feedback.
I also attached an exemplary dataset (with OBJ and texture / normalmap) to this post
Regards
Tobias
----------
For those of you who are interested more: We’re currently on Kickstarter with the project
https://www.kickstarter.com/projects/3digify/3digify-build-your-own-professional-3d-scanner
please remove this link, if considered as perky advertisement
https://www.kickstarter.com/projects/3digify/3digify-build-your-own-professional-3d-scanner
please remove this link, if considered as perky advertisement
Replies
One of the bigger challenges is finding the right asset since you can't just whip out your magic wand and make it appear.
I have attempted some homebrew 3D scanning in the past using structured light and the results were surprisingly encouraging. Having a refined solution for that would be fantastic, as indeed the price of entry can be quite low (basically the price of a low-end projector).
Are you guys going to release a rotating base - or maybe providing parts or instructions to make one ?
1:1 Asset Replication - as others pointed out a lot of studios are utilizing 3d scans to get real world assets into their pipeline from the likenesses of actors and models for digital doubles and characters to props and texture/material to live sets and environments for virtual settings. you can think of 3d Scanning as Next-Gen photography in this regard.
Archival Reference - if i remember correctly the Smithsonian institute has been investing heavily into 3d scan and printing so they can preserve their rarest artifacts by allowing 3d printed replicas be displayed as the original is safe in storage and if worst comes to worst they have the digital scan available if the original artifact is ever damaged or destroyed. product manufacturers also use 3d scan for a similar purpose where they have a virtual inventory of discontinued parts in case someone needs a replacement.
Asset Capture for Refinement - a mix between the first two points where rather than keeping the scanned asset intact it is edited for improvement. this is commonly seen in industrial manufacturing where a one-of-a-kind element or component is damaged so you 3d scan it and repair/reconstruct said component in an app for 3d printing or manufacturing. in entertainment lie VFX or game creation 3d scanning is used to get traditionally sculpted assets into the digital pipeline, for example the Cave troll from Lord of the Rings was a traditionally sculpted maquette which was scanned as a digital asset to be modeled and rigged a a 3d character. and recently archeologists and museum curators have used the same techniques to "restore" newly found artifacts that are found incomplete such as statues missing limbs or sections.
Accessory Generation - similar to the last point but instead of repairing/refining an existing part the scanned asset is used as reference or foundation to build a unique asset off of. in industrial manufacturing this is used for creating "bridge" assets meant for retrofitting two distinct components, such as making a smart phone holder for a vehicle that never had such a feature. in entertainment 3d scanning is used to get real world assets to build props off of, such as scanning a lead actor to build a costume off of. Legacy Effects is a studio that specializes in such 3d designed assets,starting with the titular mech armor of the iron man films. hobby specialists also use this technique to create custom assets for existing toy franchises,such as 3d scanning client's likeness to be printed as an action figure.
Scientific Analysis - related to the second point but rather than saving an asset for preservation it monitors said asset in (relative) real time. a recent example of this was a demo of Autodesk's Memento app in which a team of marine biologists used 3d scanners to capture a coral reef at one point in the year and then captured the same reef later in the year and compared the differences between the 3d models to determine how much said Coral has grown.
hope this helps!
Appearantly I was not utterly wrong when assuming that 3D-scanning technology seems to be of help for 3D-modelers.
My next question is targeting those who have already used 3D-scanning / photogrammetry or plan to use it: If using 3D-scanning as a tool, what would be the features you would care most? Amount of details being captured (resolution)? Ease of acquisition? Speed? Automatization? Low Price? ... ? Do you care more about capturing the geometry or the color? Whatever else comes to your mind.. Of course having everything at the same time would be ideal - But what would you really care most about? When developing our tool we would really like to target CG artists as potential users, so it is very important for us to understand how you use such tools and what you care most about.
@Pior: Nice to hear that you already attempted implementing a structured light pipeline. I'd love to see the results of that if you still have them
For instance I'd gladly embrace some limitations in the process (longer scan time and/or smaller acquisition volume for instance) if that meant that the process could be a one click solution. That said it would also be important to be able to interrupt the process at any stage to access the in-progress data if needed (for instance : being able to salvage a broken scan by still being able to access an OBJ of the first pass)
Colors would be nice, but would definitely be a secondary need since reprojecting photos onto a 3d model is trivial even without UVs when done through Zbrush polypainting - and that's something that a very large portion of your potential userbase will be able to do by themselves.
From my very limited experience I was very pleasantly surprised by the accuracy of structured light scanning, even on my terrible garage attempt
The point cloud might be low res, but I was blown away by the fact that the hard surface quality was there and even the subtle diagonal line was acquired (I hope I am not confusing my various scanning attempts here, but I *think* this was structured light. This is old though, I can't double check
My setup was very barebones - slanted photographs with a cardboard calibration background.
Now the problem as mentioned above is that while I was able to get very clean fragments of scans, the stitching process was a pain and I never ended up getting full scans out of it. (note that I didn't really have the time to investigate the process any further, so I might be biased here). I also didn't have access to an automatic rotating base.
By comparison I could get more usable results out of 123DCatch or Agisoft, because even though their output might be more lumpy and distorted/slanted, at the very least they would automatically spit out a 360 reconstructed model which was ready to open in Zbrush/Mudbox.
I remember messing around with an actual consumer grade desktop laser scanner at some point, but even that wasn't great at stitching. So if you guys can provide a rock solid reconstruction/stitching algorithm, it could really be a game changer !