As I have grown in my position as the Immersive Scholarship assistant, I have found the processes of photogrammetry and 3D scanning to be immensely enriching—though not without difficulties. In developing my skills, I have encountered a variety of obstacles that complicate, if not completely prohibit, the creation of 3D models. The most challenging model by far has undoubtedly been the fibula pictured above. This fibula, sourced from Cosa, Italy, would have been used as a pin or brooch in ancient Rome. It is one of several artifacts documented by Digital Scholarship Librarian Matthew Hunter as part of Cosa Excavations’ ongoing digitization efforts. According to the Archaeological Institute of America,

Cosa boasts a remarkable history that spans more than 1,300 years; founded on the eve of the first Punic war, it was destroyed sometime around 70 BCE and then substantially rebuilt under Augustus. Occupation, albeit in intermittent fashion, continued at the site through Late Antiquity and into the Middle Ages.[1]

Cosa’s richness as a cultural heritage site teeming with archaeological discoveries makes it a prime candidate for digitization. Its artifacts, however, present distinct challenges. The problems with capturing this fibula and other objects like it are manifold: beginning with their complexity. The fibula’s intricate, twisting design makes it onerous to capture using a 3D scanner such as the Artec Space Spider. Though the scanner is meant to capture small objects, the fibula’s surfaces are ill-defined and therefore troublesome to align in Artec Studio 14. The scanner would have immense difficulty distinguishing the top of the fibula from the bottom or, as is the case with other unsuccessful renders, fail to align the edges altogether. Failure in alignment creates pockets of negative space in which the model lacks textures or a proper mesh. I have encountered this issue several times—most notably when attempting to scan silverware from FSU Libraries’ Special Collections that was both thin and reflective. These issues can be mitigated when switching to a process such as photogrammetry, but they are not entirely eliminated.

Photogrammetry presents a different set of issues—particularly movement. Movement from either the object or the person capturing it can generate poor photos that will adversely affect the creation of a model in software such as Agisoft Metashape. When photographing the fibula, Matt found that the pin that bisects the ring would also move as he rotated the turntable. These slight, borderline imperceptible movements can cause failures in alignment and post-processing—much like those that occur during the 3D scanning process. While collaborating on the model with Marcelina Nagales, Manager of Scientific Applications at the Research Computing Center, we found that flipping the fibula to photograph its opposite side had caused shifts in the orientation of its central pin. These shifts made themselves evident after attempting to combine batch processed images of both its top and bottom. Other technical concerns included the absence of scale bars, poor lens focus, and inconsistencies in camera positioning. These irregularities generated models that were either incomplete or improperly calibrated, making them difficult to align. Despite these issues, we were able to produce decent sparse point clouds (as seen above) that adequately visualize the fibula.

Overall, I would still deem the project a success. Even if we are unable to produce an exportable model of the fibula, it will have still served its purpose as a key site for further learning and experimentation. Conducting tests on the fibula, in concert with prior digitization projects, has better informed my understanding of the procedural benefits and tradeoffs that exist between these two 3D modeling processes. I am hopeful that these lessons will prove fruitful for our next graduate assistant who will be continuing my investigations into both photogrammetry and 3D scanning.

[1] “Cosa Excavations,” Archaeological Institute of America, October 22, 2022,