Anatomy and Art

a blog by Sara Egner

3D from 2D

without comments

My plan to get a full figure out of 123d Catch has so far fallen flat.  My efforts seem to alternate between mostly outright failures sprinkled with unusable results.

Screen Shot 2013-09-04 at 11.30.31 AM

I think that the best capture we got was in the office lunch room, using the iPad.  As you can see, we still got a lot of scatter from the surrounding room.  And it’s harder to see in this image, but you can tell that there was a tiny bit of movement that resulted in a little bit of distortion in my model, Ernest’s, face.

Screen Shot 2013-09-04 at 11.38.18 AM

I think that the lunch room capture did better than the office capture near a window or the outside/ sidewalk capture because the light was flatter.  I’ve noticed that the program seems to have trouble with both shadows and also bright highlights.  Which is why we decided to try taking photos one more time at my place, with my Cannon 20D SLR camera, late in the day so that the light wouldn’t be more even.  The below images are some of what was shot, to give you an idea how we were going about this.  I went all around him, then got up on a step ladder and got shots all around from above.  I made certain to get the top of his head, since we’d had trouble with that in a previous attempt.  I got down low and shot upward to avoid any blank spots under the chin or hands.  I took closer shots of the hands and legs and feet to prevent a previous effect I’d seen where the legs and floor between them had been treated as a continuous object.


I think that we had the right idea here, but I’ve since learned that the software relies a lot more on background overlap than I’d imagined.  So while I was sure to get overlap in my shots of Ernest, I didn’t think about making sure that trees were identifiable in the background from one shot to the next, or that I include the kind of overlap on the house that would help the software align shots.

My error became clearer when I tried to crop all of my images down to just Ernest, thinking that it would focus the software on the intended subject, but my results only became worse for doing so, and as I read more about it found advice to the contrary.  If I were to do this again, I would put clear identifying objects on the ground near my model to help the software key.  And I certainly wouldn’t have let the ladder ever be in different locations for different shots in the background.

I would like to try this again sometime with better attention paid to my lens as well.  While I wasn’t using a telephoto lens, I did have a moderate zoom lens on and I’m pretty sure I must have adjusted that between shots a little.  I also wonder about how the depth of field affected my results.  Sometimes there’s only so much you can do, especially shooting a tall lean figure like Ernest.  But I’m sure that as broad a depth of field as possible would be better.

Ideally, I’d have something like 60 cameras set up all around to cover all angles and on different colored tripods to help the software key each shot as it makes it’s shape, all firing at exactly the same time.  I could capture action shots with a set up like that, and skip the race against the setting sun entirely.  But I still think there’s possibility to get somewhere with what I do have.

Anyway, I had to move on to another method of doing things for my work project, but my interest in figuring out this technology remains.

Written by Sara

September 4th, 2013 at 10:42 pm

Posted in Uncategorized

Tagged with

Leave a Reply