|
The objective is to find a good middle ground in quality and price of scanning humans. First of all this is an interesting "low cost" approach I stumbled upon: I did my own experiments with some external raw photogrammetry data and measured around 13 cameras needed to get a relatively good face and moderate head scan out of them (by choosing specific photos with various angles of that dataset..so its under ideal conditions). Both would not capture pore level accuracy, but the head shape can be captured with relatively good textures. Acquiring the cameras wouldn't even break the bank: You can get used Canon 600D for 240 bucks which would includes the battery adapter a kit lens, memory card and maybe even a cheap ball head. Of course you will still need stands and lights, including modifiers, USB cables, and hope that your custom written program can cope with that amount of cameras. This is still outside my current budget (and justification), but maybe not outside the range for someone who wants to invest in such a setup professionally. But to be fair, it is still far less than the 100k and more, what a full commercial 80-100 cam rig would cost The lidar shown in the video on the other hand as beyond my reach, and will ever be. I've accepted that that I never will squire a faro or get hands on experience. But lets be more economical....and unprofessional. An alternative would be using a revopoint miraco (remember those are practically useless in daylight..but then again you would try to scan in darker environments anyway to avoid catching environment light as much as possible) or a shining3d vega (which I don't have, but they claim being able to scan in sunlight). Both would cost less as those 13 cams, even much less if you buy used. And from my experiments you are on par in mesh quality in comparison to the limited photogrammetry setup. The additional advantage would be that it is extremely portable, has fast setup time (no need to setup cams and lights) and gives you instant wysiwyg. For high res textures an additional camera with polarisers for cam and ring light would be needed. So you can shoot some additional photos to project-paint them on the mesh at later stages. The best thing of this setup is, that its dual usage: You can use the cam + ring light for texture photography and regular photogrammetry. Same for the portable 3d scanner. (To a degree). So the justification level is much lower for an investment in such a setup. As for full body scans (and I did not test this yet) things get trickier. Personally I would use my revopoint range, and hope for the best that it does not loose tracking. (Making two separate scans for front and back, I would not trust it to cope with a full 360 turntable scan). But revopoint discontinued it...so I'm not sure which of the current available scanners offer a similar scanning volume at a low price. The idea is to capture as much as possible to avoid scanning errors by unwanted movement of the subject. But again there some used / clearance sales out there. Guess Revopoint want to sell more of their Trackit sets. But this is, with 4k price, definitive outside my range. So yeah, that’s my current train of thought on this. I'm posting this, since I currently have other projects I want to finish first, before diving into this subject myself. Maybe a phone scan of the body is enough for your stuff. Maybe you enjoy traditional head modeling on the basis of some photos....yes there are cheaper methods, and there definite BETTER methods...but personally I think this method is located at a sweetspot in price, speed and quality. So sorry for not posting a full fledged tutorial, or an example. I'll do it, at least thats the plan. but don't expect beeing this year or the next. |
