One more scan. It was generally a test and a phraze from my friend: “Hey, can you scan me?”… Scanning itself was done manually and took about 1-2 minutes just plainly at the yard with my friend standing on asphalt. But to get the result it took me literally about 12 hours if not more. It was a curiosity leveled to “I’ll get this scan done anyway!”.
“What took so long?!”. As I’ve seen (and you might too), theese kind of scans have a quite ammount of imperfections like a lot of double surfaces of that lovely drifting tracking, so when you try to mesh it, you’ll get a double surfaced nose and things alike. Sure, there might be a better method of dealing with those, but I’ve just gradually filtered in CloudCompare one bug by one. Had to manually add some parts like missing fragment of the hat, remove holes in the ear and in some more places and so on.
And considering theese poisonous frames that brakes down consistency I have some feature to implement into RevoScan (a merely easy to make one) that can help dealing those at the Raw level just before even fusing the vertices. I think will post that in an appropriate section, but I have some fear nobody except PUTV really read those…
Anyway… One more thing that made this one very long, I wanted to get 11 level of octree depth in CloudCompare. It lets to exctract some more details even they don’t benefit to the appearance that much. But the “11 level” of it make it way more demanding to the point cloud quality. And when there are always 2-3 layers of surfaces in fact, it might pick up theese and make resulting mesh very buggy. Don’t ask me why I haven’t ensured those overlaps were removed - maybe they were deep enough to eat actual details like ears or the hat thickness, I just don’t remember. Maybe that was just me making it the wrong way. But the general idea was simple - make it 10 octree depth and it will be just OK, make it 11 and have bugs over bugs.
This how it did look in RevoScan. To the left is the original scan result in point cloud form, to the right is a mesh generated at RevoScan at maximum quality (6 it was AFAIR). Notice missing parts at the hat and between legs.
This shot shows why there was no way to directly sew up those missing parts of the hat and trousers… And how much problem that hat did make )
To solve this missing parts’ issues I had to design them manually. In general, they looks like this. This is a stitch for the trousers for one side:
If to look at all those “manual parts”, the look this:
For example, I needed to remake a handle of the rucksack and a sticking out strap of the hat.
The more or less end result I had to finish up in ZBrush since there were still a lot of stitches of manually deleted subsurfaces, so I needed to remove those. Apart from that it is still a scan and not a ZBrush work. The result look like this:
The details are actually way nice… And yes, I love that ZBrush shading too.
One thing that is stil unknown to me, would the stabilizer really help with tracking drift and loosing of alignment (since it costs me a lot of time dealing those) or the difference would be too marginal to really bother. I have some idea of second hand stab that can be used with Range, but, lets say, it not my primary concern now.
BTW: Mini was sold out for good. Yes, there are some things you literally can’t do without it, but it takes so much efforts to get a result, that you think twice do you really need to make those. It should be the “little helper” that you always can take out from the shelf and make a quick scan, but this thing actually demands worshipping it, so it goes. Interesting, there is a company that seems to have some agreement with Revopoint that Revopoint will give them Minis and they will use their own software with it. Will this make things better or not, would be interesting to know, but that’s more of “it would be interesting” and nothing more.