We are a group of students from the Master in Digital Humanities of the Karl-Franzens-University of Graz, who attended the seminar module “3D Reproductions in Cultural Heritage” (led by Ass. Prof. Chiara Zuanni). The objective of the module was to reflect upon the possibilities of 3D reproductions and their usefulness in the context of museum exhibitions and museology in general.
We were introduced to the creation of 3D reproductions of cultural artifacts, and the increasing importance that this practice, alongside technological innovations, e.g., in the field of virtual reality, can have for the fields of Museology and Archaeology, and for the preservation of cultural heritage. First, a theoretical introduction to 3D reproductions covered some considerations on the questions of authenticity that pertain to this field, as well as the possible ethical implications that have to be taken into consideration when doing this kind of work. Subsequently, we used photogrammetry, i.e., the reproduction of a model from a series of photographs, in order to create 3D models of selected cultural heritage objects. In this post, we aim to give a very brief impression of the first attempts at 3D modelling by a group of complete novices in this field. Finally, we uploaded our models to Sketchfab and designed a small videogame including these models, in order to experiment with the dissemination of 3D reproductions and experiment with different engagement strategies.
Our step by step process of creating the 3D Models
Our work focused on casts of ancient sculptures displayed in the Gipsmuseum of the Archeology Department at the University of Graz. This museum was founded in 1865 and it exhibits casts of the most significant sculptures of Classical Antiquity, as well as a remarkable ceramic collection. We used Agisoft Photoscan to produce our models.
Organised into several two-person teams, with one student responsible for taking the photos and the other for monitoring the process, we spread out in the galleries to choose the objects for our 3D models. Each team took photos of 2 to 3 objects, the minimum number of photos that should be taken for each object being 30.
At this point, it is important to note that more photos do not necessarily translate into a better quality of the 3D model. This does not mean that one should aim for a small number of photos, but rather that working precisely on each photo, paying attention to the lighting and the coverage of the object, has more impact on the outcome than a big number of photos taken. For a satisfying result, it is absolutely necessary to capture the object from every possible angle and these photos should overlap to a certain degree. If the object is not completely recorded or the overlapping degree of the photos is not appropriate, the photogrammetry software might interpret missing parts as holes in the finished 3D model. In addition, photos should obviously be in focus and, since we worked without a rotating table and markers, we had to make sure that the object was not moved or rotated while taking the photos.
As a side note it should be mentioned that the background can be masked in the photogrammetry software, which meant that there was no need to worry about it while taking the photos.
The masking step might have been the hardest part of our project. Not because it was difficult to accomplish, but because masking the photos was quite repetitive and took a few hours of work. Since the background of a photo is not part of an object, it has to be removed or, more precisely, ignored. Therefore, one has to tell the software which parts of a photo make up the background and which make up the object. This can be done by masking those areas appropriately. At the end of the masking process, the user ends up with photos that clearly delimit the object. As already mentioned, this is a task that mainly requires time, a good amount of patience, and maybe a bit of practice with the selecting tools.
The Workflow of the Software
At the end of the masking step, we were only a few final steps away from our finished 3D models. In Agisoft Photoscan, we followed the standard workflow, in four parts, which is a routine that is done entirely by the software.
The first part is the alignment of the photos, in which the software calculates the individual camera positions of each photo. The second part is the creation of a dense point cloud, and we could select the level of quality and other parameters prior to its construction.
The third part is to build the mesh of our models, which can then be exported to other analysis tools for further editing or working with it. For example, we used Meshlab to attempt to correct and improve some of the models (e.g. when the masking process had not been done accurately and we needed to remove further parts or, vice versa, when we need to ‘fill’ some of the holes. The fourth steps in Agisoft allowed us to add the textures to our models, and afterwards we went on to build tiled models (as enabled by the professional edition of Agisoft Photoscan).
Dissemination and engagement with the models
Once we had completed the models, we reflected on how they could be used in a museum context to engage with visitors. Therefore, we decided to upload them on Sketchfab, a process that was very simple and straightforward, and allowed us to showcase our work. Not only could we describe the objects, but we could also add annotations to highlight specific features, and we could view them in Virtual Reality with Oculus Go. At the same time, we also experimented with Unity3D, developing a small videogame including our models. This brought us to discuss how museums could use 3D models to disseminate their collections, develop their engagement strategies, and reflect on the museological implications of these new methods of reproducing and sharing objects (i.e., in terms of authenticity, ethics, and curation and preservation practices).
We agreed that while photogrammetry is relatively easy to learn, producing good models requires experience: from the quality of the photos to the choice of settings in Agisoft, there are a lot of areas in which our work could be notably improved with more practice. Hence, the modules we created and uploaded on Sketchfab represent the first attempts at 3D reproductions by complete novices to this field. For example, a better camera, more attention to the light, and more attention to the coverage of the objects when taking photos, would certainly help us to improve the detail and quality of our objects. However, these models still show the potential of photogrammetry for disseminating museum collections.
Uploading the models on Sketchfab and working with Unity prompted us to reflect on how our models would compete with commercial products, for example by the gaming industry, which are produced by bigger teams with more financial resources. We thought that, in this moment, applications of VR and gaming by museums cannot compete in detail and design with industrial ones. However, solutions such as the ones we experimented with could be useful in prompting more curiosity about heritage objects and reaching new audiences. 3D models of heritage objects could therefore be a fruitful way of disseminating the richness of museum collections and the expertise of curators in new ways.
This article was written by Fruhwirth Alexander, Galka Selina, Hartleb Christoph, Sonnberger Jakob, Stoff Sebastian, Tscherne Niklas, Zuanni Chiara on behalf of the ‘3D Reproductions in Cultural Heritage’ class (Zentrum für Informationsmodellierung – Austrian Centre for Digital Humanities, University of Graz)