My Experience 3D Modeling Birch House

Introduction

This week in DGAH 110: Hacking the Humanities, we learned about various techniques for turning a series of photographs of an object into 3D models. Specifically, we learned about two mains approaches for Photogrammetry: using cloud-based platforms such as Polycam or Scaniverse, or with local computation using Agisoft MetaShape. I attempted to model Birch House, one of Carleton College’s new houses in the Lilac Hill housing development, to varying degrees of success. The models I have obtained using both of these methods can be found in the section below.

Polycam Model

This model was created using the iPhone app Polycam, using their cloud algorithms to reconstruct a 3D model based on photos. As you can see, this reconstruction is far from perfect – it contains many missing portions of the mesh, mainly areas where the camera was not able to get a good view. Additionally, the reconstruction algorithm missed one of the corners, resulting in a large gap between two walls that connect in reality.

MetaShape Model

This model was created using the Agisoft MetaShape desktop application for Windows, through a process of photo alignment, building a mesh model based on the aligned points, cleaning the mesh, and then refining the textures on the mesh model. As you can see, this process was not very successful – even less so than the Polycam model shown above. One main reason for this was the fact that it was hard to photos located a fixed distance away from Birch House, due to the surroundings of the house. Birch House is located in close to proximity to other houses in the Lilac Hill development, as well as to an area with many thick shrubs, especially on the rear side of the house. This prevented me from successfully getting photos with the whole house in view from a fixed distance, limiting the number of corresponding points that MetaShape could pick out. Had I been able to take sufficient photos in a complete circle around Birch House, I think that this method would have resulted in a good looking 3D model.

Reflection

Attempting to build a 3D mesh model of a building captures many details that do not successfully come through in photographs alone. For example, models convey spacial relationships between different portions of a building, allowing viewers to more successfully understand the purpose of a building. Viewers might see a central entrance to a building, with many levels consisting of windows towards the back of the building, and understand that this is a house with dorm-style living. Finally, modeling can also capture important details about the construction materials and architecture style of a building that does not come through in a photograph. This can help inform things about the time period that a building was built in, or the location in the world.

Additionally, one thing that I learned during this lab is that the task of modeling itself, regardless of the outcome, is a valuable process that teaches us as digital humanists much about the spaces that we inhabit. First of all, it gives me a much greater appreciation into the thought behind designing buildings and living spaces on campus. The Lilac Hill development at Carleton was clearly thought out with students in mind – from the bike racks in front of all the houses, to the porches and the placement of windows throughout the houses. It also makes me more appreciative of the architecture and design style of the houses in Lilac Hill. These houses visually differ much from other buildings on campus, and it is interesting to look at the history of Carleton through the architecture of buildings. Buildings more central to the original campus building, Willis Hall, have a much older architecture style, and as an observer moves outward to the edges of campus, they will be transported through time as more buildings were constructed throughout the decades. Small details such as building style are things that the active process of building modeling helps us as digital humanists to understand.