Hey,
 
I've setup a light tent (for light diffusion and minimalization of reflections/specular highlights), followed the directions on taking photos, applied a grid to the object, and took 18 photos (9 of the top and 9 of the bottom, at 40 degree intervals).  There is sufficient information and points of reference (via the grid) to convert the entire object to 3d.  And, to boot, I have specially chosen a set of points for calibration so that each (and every) photo has at least 3 references on it (some with 6 or 8), carefully tying together those from the top and bottom for continuation.
 
Yet, when it comes to calibration, only six or so of the photos get the "green light", all of the references have 0.000 for residual value (obviously not because of accuracy), and points get misreferenced on some sets where they aren't even visible.  Can ImageModeler not handle 360 views for full modeling?
 
This is my second attempt at this object, this time being thoroughly methodical, and still can't get a model out of the darn thing.
 
What am I doing wrong and how can it be corrected?
 
Robert Templeton

Just to add:
 
The object is a spoon, which is why I glued a grid to it.  There are no distinguishable reference points otherwise, all curves.  So the grid gives me precise, exact, easily distinguishable, direct reference points.
 
I loaded just the top set (9 images) into IM and added points all the way around the periphery (not planar by any standard), down the center from stem to bowl (nonplanar and not following the periphery points) and some inbetween on the bowl.  Eh hem... That's 23 points.  And it still could not calibrate properly.  Only cleared one (1) image and a couple of locator points.  Believe me, placement accuracy is not an issue and many references between photos (3, 4, 5, 6).  How many does it take!?
 
What's up here?  Seems to have extreme difficulty with concave surfaces.  Is there a return policy if this doesn't get solved?
 
FYI: Same digital camera, same focal length, same lens, no flash.
 
Robert Templeton

Hi Robert,
 
As Greg said in an other post, it is preferable to calibrate dynamically :
calibrate first 4 images, then add the other shots one to one in the calibration.
 
Generally, it is usefull (especially in case of organic objects) to add another object with "good" visible features and good 3D information (i.e. points in all dimension) near your organic object (ex: a cube).
 
The grid will be more helpfull to place modeling points, once it is calibrated.
 
Stef

Hello Robert.
 
To add to Stef's comment,
 
Imagemodeler works on color pattern recognition.  If your grid point are all the same color, there's a good chance it's getting confused.  I think you'll find if you use different colors, you'll come up with a much better result.  I had this problem when first starting out.  No matter how many times I reshot my object, it wouldn't calibrate.  Switched to a multi-colored solution, and it worked the first time.