I've never made a blog post before.. but I wanted to keep going on this project and be consistent with it.
I'm
posting this here for my own sake but also a way to be more involved in
our industry community, learn new techniques, and hear other artists'
perspective or advice, if you're so inclined to give it 😁
I
recently got into photogrammetry and photo scanning. I use a Nikon D3200
which I got a few years back. It seems to produce pretty good images,
if you can get your settings right.
Anyway, I set out to scan a
number of objects, all of which (3-4) came out with photo alignment
issues. My images were probably too dark/too narrow depth of field/not
enough shots.
Now, I said, with this Pikachu stuffed animal, I'm
going to get this right! Who doesn't like Pikachu? So all in, I have
about 400+ images. I took about 350 to start, then realized I probably
needed bottom images, and more detailed shots to clear up the
discrepancies in his tail/ So I took about 56 more, under preeetty
similar lighting conditions.
Overall,
the issue I am running into is that the two components, once in Reality
Scan, do not align. It's not that there are photo alignment issues this
time. It's that the objects themselves do not align, and I'm not sure
if Reality Scan will be able to merge the point clouds because of this.
Yes, that is an image of Pikachu's butt. But I felt I needed that side to get a complete and full mesh!
I'm
wondering if I need to re-shoot, or if I can Set Ground Plane to adjust
the pivot/transforms/location of the butt version component (for lack
of a better name). Hopefully this makes sense. Please feel free to
comment and let me know if you have any ideas! Working on solutions to
move this one forward.
Thank you,
& cheers!
Adam Rice
Check out my portfolio at adamriceart.com. Password to the Games page available upon request!
Shout to Clinton Jones's (pwnisher)'s Youtube tutorial found here:
Plush is probably not the best choice for (detailed) photo scanning, as the deeper parts between the hair are obscured pretty quickly when you change your position and you'd need a crazy resolution to scan the actual hair. The overall falloff/fresnel effect caused by this might not matter that much, but could still play a role as well. Also, handling a plush toy will change some of the hair, and placing it differently will likely lead to gravity moving some of the limbs slightly and of course the overall lighting situation will change if you orient the toy differently. You have a pretty strong bounce light from the table, for example.
It also looks like you have too little overlap of the bottom piece with the rest to set meaningful anchor points, so you might end up with two separate meshes you'll have to roughly align and combine in the end and you'd have to fix the texture anyway. Not sure if that's worth the hassle.
You could try placing it on a stick or something if you manage to make it stable and suspend it that way to get all photos in one go, but you probably won't get a better surface than what you already have. You could even try to go with a lower resolution and less pictures, going for bigger details and maybe getting a smoother surface that way. But an object with a matte and relatively smooth, but detailed surface would be a much better choice.
Noren said: "an object with a matte and relatively smooth, but detailed surface would be a much better choice."
Agree, I've never scanned anything but it seems like a fun learning project.
Here are my 1st thoughts about what to do with the complications of this particular object, it doesn't look too "expensive", so I'd say get another one and "shave it till you get to a decent amount of fuzz" of if its possible zero fuzz.
Then Rescan for fuzz height? & figure out the steps to "reproducing" the fuzz later on?
i tried, again never did it, seems fun, tried to add a solution that might work, all the best with this adventure, if you return with anything i'll check it out.
Replies
Hi there!
I've never made a blog post before.. but I wanted to keep going on this project and be consistent with it.
I'm posting this here for my own sake but also a way to be more involved in our industry community, learn new techniques, and hear other artists' perspective or advice, if you're so inclined to give it 😁
I recently got into photogrammetry and photo scanning. I use a Nikon D3200 which I got a few years back. It seems to produce pretty good images, if you can get your settings right.
Anyway, I set out to scan a number of objects, all of which (3-4) came out with photo alignment issues. My images were probably too dark/too narrow depth of field/not enough shots.
Now, I said, with this Pikachu stuffed animal, I'm going to get this right! Who doesn't like Pikachu? So all in, I have about 400+ images. I took about 350 to start, then realized I probably needed bottom images, and more detailed shots to clear up the discrepancies in his tail/ So I took about 56 more, under preeetty similar lighting conditions.
Overall, the issue I am running into is that the two components, once in Reality Scan, do not align. It's not that there are photo alignment issues this time. It's that the objects themselves do not align, and I'm not sure if Reality Scan will be able to merge the point clouds because of this.
Yes, that is an image of Pikachu's butt. But I felt I needed that side to get a complete and full mesh!
I'm wondering if I need to re-shoot, or if I can Set Ground Plane to adjust the pivot/transforms/location of the butt version component (for lack of a better name). Hopefully this makes sense. Please feel free to comment and let me know if you have any ideas! Working on solutions to move this one forward.
Thank you,
& cheers!
Adam Rice
Check out my portfolio at adamriceart.com. Password to the Games page available upon request!
Shout to Clinton Jones's (pwnisher)'s Youtube tutorial found here:
The Ultimate Guide to 3D Scanning | FULL WORKFLOW
And also to William Faucher's tutorials (I've practically watched all of his), specifically this one:
How to 3D Scan All Sides of Objects - RealityCapture
Also, handling a plush toy will change some of the hair, and placing it differently will likely lead to gravity moving some of the limbs slightly and of course the overall lighting situation will change if you orient the toy differently. You have a pretty strong bounce light from the table, for example.
It also looks like you have too little overlap of the bottom piece with the rest to set meaningful anchor points, so you might end up with two separate meshes you'll have to roughly align and combine in the end and you'd have to fix the texture anyway. Not sure if that's worth the hassle.
But an object with a matte and relatively smooth, but detailed surface would be a much better choice.