A (Flawed) Final Project Data Visualization

Our data is unique in that it originates from a physical object housed in the Gould Library Object Archive here on campus, specifically, the 1940s era Adelphic Literary Society paddle. The technical aspect of our project is (or was supposed to be) a 3D model of our chosen artifact using photogrammetry via Recap Photo, but our initial attempt was unsuccessful. First and foremost, we underestimated how precise the setup needed to be in order to produce an identifiable and accurate model, resulting in unusable photos (and far too many of them) that ultimately produced an unsuitable model. 

Before we discuss what went wrong in our first attempt at doing photogrammetry on our own, let’s address why we chose a 3D modeling format. Our project is centered around the hidden history of Greek Life here at Carleton in centuries past, namely the Literary Societies. The archives have a number of authentic and well-preserved objects used by (in the case of the paddle, we have photographic evidence of it being used) the Adelphic society, and were eager to help us try to make a model of the paddle as a proof-of-concept for digitizing other physical archive materials. With respect to its function in our project, we’re using photogrammetry in a more exploratory manner, making a digital copy of a physical archive object that can be observed by users.

With respect to our initial attempt at modeling, we took around one-hundred-and-thirty photos, which went over the maximum capacity of the educational license object-based-model option. Before making the model, we used iMazing converter to change the file type of our photos from .HEIC to .IMG. Unfortunately, our photo collection contained a fair amount of background contamination—that is, a lack of consistent background (behind the object) which confused the algorithm. This led to random pixels from the background being matched in the final product, which only contributed to the model’s jumbled and confusing appearance.

Our second attempt, which is currently rendering, attempted to resolve the issues of our first attempt by using multiple backgrounds and taking photos that included nothing save the paddle and the colored backgrounds, of which we used two—one for each side of the paddle. Because we only have the model from our first attempt finished so far, the new model will not make an appearance in this post. However, you can see below what happens when you do not carefully vet your photogrammetry setup: