From the Autodesk web site: "Autodesk® Stitcher™ Unlimited photo stitching software offers professionals and enthusiasts easy-to-use tools for creating high-quality panoramas. From dramatic backdrops for film and game productions, to stunning images for artistic display, Autodesk Stitcher Unlimited panorama software can automatically assemble a wide variety of images."
Since Project Photofly is brought to you by the same team that created Autodesk Stitcher Unlimited, some users are under the misimpression that Project Photofly is similar to Autodesk Stitcher Unlimited. Although Project Photofly does stitch pictures together as part of creating a 3D scene, it does much more. It constructs a 3D model by generating point cloud data. With the Photo Scene Editor, you can even add reference geometry based on the points. All of this can then be imported into applications like AutoCAD, Inventor, or Revit. When taking pictures to use with Project Photofly, you should keep this in mind. In fact, Autodesk Labs VP, Brian Mathews, addressed the differences between Project Photofly and generic photo stitchers in a recent reply in the Project Photofly discussion forum.
Many people are influenced by their experience with the creation of panoramic images from a set of smaller images. Autodesk has a product that does panoramas called Autodesk Stitcher Unlimited, and there are several other products that do something similar. In a panorama creation scenario (like with Autodesk Stitcher Unlimited) the most optimal way to create the photos is to place the camera on a tripod in the center of the room and take multiple pictures where you rotate the camera around an axis that is as near to the plane of the camera's "film" as possible. Each photo should overlap with the photo to the left and the photo to the right. As such any particular object in the room is generally only seen in either a single shot, or at most two shots (where two photos overlap from left to right).
When using Project Photofly and the Photo Scene Editor, we are not making a panorama. In this case we need to see every object multiple times (3 or hopefully many more) from different camera locations. So rather than keeping the camera at a fixed position and spinning on an axis, the camera needs to move so we can create triangulations on the objects. Rather than just overlapping pictures on the left and right side by some amount (whereby specific objects are only "seen" once or twice) we need each object to be seen multiple times from multiple camera locations. From that we compute the camera positions, and from that we compute the 3D coordinates for the objects we see.
For this reason a wide angle lens can help so that there are many features in each shot that can be successfully matched to lots of other features found in other shots. If you use a very zoomed-in view then there are few "global orientation" reference points to work with.
So in summary, you can do what you want, but if you take your pictures incorrectly (too many object occlusions, poor photo scheduling, etc.) you will get poor results. Don't think like "panoramas", think like "triangulations."
You can get started with your own triangulations at
Sharing information is alive in the lab.