how I learned to stop worrying and love the mistake

By Tyler

As my digital history course draws to an end it occurred to me that I never talked much about the process of trying to create 3D models for our group project. This may have been because I never personally had much success but here are my quick thoughts on the process anyway. Hopefully my mistakes will prove helpful for others facing similar issues.

My attempt at 3D modelling has not yielded any worthwhile results. I have however learned a number of things about 3D modelling in general as a result of my toil. I attempted to create a 3D model of a DC-8 aircraft model from the Canadian Science and Technology Museum. The two different software I tried to use were 123DCatch and Photosscan. 123DCatch is the more simplistic of the two but both work by importing a collection of pictures of the object, identifying points of reference in the photos and then constructing a mesh of the object from those points. Using 123DCatch does not involve more than just uploading the images and hoping for the best, which in my case did not work out. I moved on to Photoscan which allows for much more fine tuning of the process. All of the models created from simply uploading my photos yielded some sort of strange representation of the space that I took the pictures in. Neither program was able to determine the focus of my photos and as a result tried to model a variety of different things, which moulded together and were basically useless. Photoscan produced slightly better results but the plane was still hopelessly fused with the table it sat on for photographing. Using the masking tools of Photoscan I was able to isolate the plane but the using rough outlines still meant the plane was fused with the table. It had become clear that one of the major problems I was have was that the table and plane were far to close in colour and as a result the program had trouble distinguishing between them. I then produced very close masks that completely isolated the plane from its environment. Unfortunately this did not yield any better results, it created a blob with multiple wings perturbing from everywhere and randomly scattered sections of plane throughout. I believe that by isolating the object as much as I did the program lots all of its spatial reference and as a result interpreted every photo as a unique object with little to no overlap. I think the ultimate solution to my problems was to take new pictures of the plane. Ideally the object should be photographed in a background that contrasts with the colour of the model. It would also be helpful to create points of reference within the back drop since the smooth uniform nature of the object does not seem to offer enough points of reference on its own. This is just my theory on how to make a more successful model of this object and unfortunately I did not have an opportunity to test it due to time constraints.

Source: Sinclair