What’s the deal?
Photogrammetry has been discussed a lot in recent years, especially inside of heritage and for good reason, it’s a straightforward and very useful way to capture archaeological material and communicate data. Yet, something that isn’t discussed as often, are the potential difficulties in taking 3D models that are created through the photogrammetry process further than just becoming a singular, stand alone asset. This is often because there is a breakdown between photogrammetry as a ‘technique’ and the wider CGI pipeline. To freely be able to make use of 3D assets like those produced by programs like Agisoft requires a great deal of knowledge in processes such as hard/soft surface modeling, texturing/unwrapping, rendering and retopology.
Why rework them at all?
There are multiple reasons that you might want to rework Photogrammetry:
- You might have multiple scans you wish to combine into a single object.
- You might want to clean up the dataset (Sort out the mesh, fill in missing information, work on the texture).
- You might need lighter models to work with due to software constraints (Game Engine assets, big scene creation/scattering).
- You might want to take/edit elements from the photogrammetry (textures, pieces, information) and have it drive other models either through texture, or geometrically.
- You might want to augment the model (Work on shaders/adding information to the asset like dirt/dust/rust/mud etc) or generally just set up a base level of control.
So what needs to be done?
Well this is the difficult and fiddly bit and like most areas of the wider CGI pipeline there really isn’t any ‘right’ or ‘wrong’ way to do it. There are countless programs, techniques and workflows that you could use, so the way to actually do this process can vary, but in short it requires a fair amount of knowledge, especially in relation to modelling, material creation, unwrapping and rendering (especially to texture/baking). Essentially, for those without this background, much of this is similar to what you might see in a traditional computer game asset creation workflow, and some of it is summarized below in the short timelapse video:
What the above shows is the process from start to finish of reworking photogrammetry into a more manageable asset that can then be manipulated at will by the artist and used as an economical 3D object for a wide range of new purposes.
Why was this needed on the Invincible?
Mainly it was required because there were a number of difficulties in obtaining the datasets:
- Conditions for the dive enforced very close up photography to capture the information due to visibility issues.
- Dive constraints and time issues meant that much of the content captured was fractured into sections.
- Varying visibility and other issues also caused differences in the final Photogrammetry.
- Not all of the site was surveyed, so other references were need to fill in missing assets like timbers/iron knees etc.
- To build up a picture of the whole site site other input data was needed (Multibeam/site plans etc).
The above are just a few of the issues people might encounter when trying to capture photogrammetry of more difficult archaeological sites. Obviously, working underwater poses very clear and often unique problems, but issues such as access, level of detail, site size, environmental conditions and other variables, can be universal. Likewise, perhaps it’s just a series of bad photographs, the lighting was problematic on that particular day, or things are out of focus or noisy and time is short and there is no time to reshoot. The ability to be able to do something with these data sets is important in providing the flexibility that is often required, but missing from 3D capture projects in heritage.
Creating the wider resource
The image above highlights the rough workflow that was developed on the Invincible and this was ultimately chosen to fulfil three main goals:
- Create a long-term resource that is useful to the project.
- Create outputs that represent the efforts in collecting the data by the team
- Provide an experience to the general public that is as accessible as possible.
The shots above show the benefit of being able to economise the photogrammetry to the point where it becomes more than just a singular stand alone object and can be incorprated into a much larger scene. 3DS Max tends to not play nicely past a certain point in its viewport performance, especially if you are introducing very dense objects into the scene, so having clean models you can work with easily is imperative, especially if you are also working with very large textures and trying to fix issues in the UVW maps for the textures and a whole host of other general housekeeping issues.
An example of the final outputs for the photogrammetry in their edited state can be found in the Tour. Below are a handful of rendered shots from the final models.
Why is the above important?
Something that has become clear during the process of creating these tours and reworking models especially from underwater projects is that the archaeology most in need of recording, either due to bad conditions, changing circumstances, or just access difficulty requires the most amount of work and skill. Being able to create perfect photogrammetry models of objects in collections is a perfectly valid exercise, but often it is the sites that are screaming out for investigative archaeological techniques, paired with highly skilled practitioners that are the ones in desperate need of digital interpretation and recording. It is hoped that through exercises like these web tours the benefit of many of the wider CGI pipeline techniques (and other multi disciplinary skills) can be demonstrated, because without the ability to apply his knowledge, many of the resources that now exist for the Invincible Team wouldn’t have been facilitated.