Graphics and Archaeology: Interpreting the Past

Snowbird, Utah
20-23 May 2000

Panel:

What does Technology Have to Offer In the Future?

Presenters:

• Paul Debevec
• Marc Levoy
• Lon Addison
• Philippe Martinez

Presentation Resources:

• http://www.debevec.org/
• http://www.cs.berkeley.edu/~debevec/Probes/

Summary:

• Scribe's note - this session record is an approximation of the lively discussion.. Elements below are in chronological order and position presentations are interleaved without headers.

Paul started the panel. How do you capture an object? What have you modeled? What have you not modeled?

• Current approach
  • scan the geometry
  • take photographs under various lighting conditions
  • derive spatially varying reflectance properties
  • render under novel illumination using computer illumination
• Problems with the Current Approach
  • [source] geometry can be fuzzy [not smooth]
  • surfaces may not reflect light adequately
  • reflectance properties can be arbitrary; some areas will be hard to reach
  • rendering is expensive; not necessarily correct
• Solution - record the Reflectance Field of the artifact
  • directly capture how the artifact responds to a dense array of basis illumination conditions as viewed from many positions
  • advantages - no models required - render directly and correctly from data
  • disadvantages - slow capture (?), lots of data
  • note - we want to record not only reflected illumination, but how the artifact responds to light
  • note - it is more difficult to recover the geometry - this only works for pristine artifacts

Paul went on to detail the Light Stage, version 1.0. This novel capture platform allowed the data capture of subjects from which image-based rendering with reflectance functions could be generated. Using a fixed viewpoint with light source information captured from all directions, it is possible to apply a normalized light map through the reflectance function for the view point [dot product of reflectance map x reflectance function = pixel from that angle].

Marc next presented a position on measuring and rendering light fields and BRDFs. Their approach involved a spherical light field camera (4 degrees of freedom). A video light field camera (possible future replacement for holograms) was in the works, part of the Standard Immersive Television project. He also covered very high-resolution laser scanning for data acquisition and the simulation of metallic patinas and fluid flows.

In answer to a question about holograms, Marc thought that Russia had some of the best implementations right now.

Philippe next spoke up on behalf of archaeologists. He offered:

• what technology should offer us
• multimedia approach has been very interesting to us thus far, especially when incorporating sound
• all of these ways of managing and creating our data, new ways of analyzing our information, it should force us to change our way of thinking about our problems
• the ability to capture data about all artifacts, in addition to most prominent ones, is very important
• we complain about expense, that we need to have things that are very power effective for remote site acquisition (power is unreliable in the field)
• our aim should go beyond just the gathering of information for ourselves, we should be able to pass our findings to our colleagues and others (edutainment)
• we have to be able to bring back to life all of the elements that contributed to an artifact (light, sound, smell, etc.)

It means more than just reliable equipment, we have to deal with a huge volume of information. We need techniques that not only capture data well in the field, they should also work well for our colleagues back in their offices - large amounts of data need to be accessible.

Duncan noted that archaeology lost part of its popular basis when we went scientific with carbon dating. This new technology is giving us a chance to re-democratize the data. The encouraging thing is that we're rediscovering our links back to the people/public.

It was noted that archaeologists don't really care about how a result was produced, but they also rely on the credibility of the result itself.

Lon added to the conversation that there are lots of other areas that we need to think about. To that point, we had only discussed about surface technologies. Sub-surface gathering (such as medical imaging) was also very important to be considered. QuickTimeVR and video walkthroughs are important for non-rendered results. Having live linkages are important in the field (two-way satellite dishes). Processing volumetric data is a big challenge, too.

David asked what are the grand challenges for analysis (more than just artifact acquisition)? What problems would we like to solve? Lon answered that we've focused on built-heritage... there's also natural heritage. There should be tracking technology for a biological ecosystem.

What should we be prosing as research projects for the future? Ruth answered that different archaeologists have different ideas of what they'd like. She's interested in perceptions of the past... juxtaposition of different conditions is also very valuable. Others may not be interested in these processes. Furthermore, there is fire tracking (arson research), contact traces (damage cause by two materials in contact now) and modeling other long-term processes on the computer. She did express nervousness about decision-making models in driving social negotiation.

Paul asked how can technology help move the research forward? Artifact capture seems well along on its away. Donald offered hypothesis testing is valuable. If a building was here and it burned down, would it be like the archaeological record? Accelerating decay and damage (fast-forwarding time) could be helpful. Lon agreed. Yes, accelerating time on the Colosseum would be very interesting and important.

Anne commented that it was difficult for her to envisage the future when she didn't understand what the full capacity of the technology could do. Also there was some concern about the fact that some of these things were escaping her control. She likes to think about things before jumping into them. It is also a Western-centric point of view that we assume that the artifact was created for an audience - archaeologically, there are many elements where it is the process that is important and not the artifact itself.

Duncan added that we're trying to understand human behaviour here - we're focused on process here... but estimating only from what is left behind is only part of the story. He's trying to understand the Medieval mindset - understanding cultures through a process.

Ruth commented that she would love to be able to express how she gets to her interpretations, not simply just from the artifact to interpretation. She wants to get into the prehistoric peoples' intentionality, to get into their minds, their fears, their intentions. To be able to express her interpretations is to express her imagination.

Marc noted an analogy here. In computer graphics there was a big period of tool building for visualization. It took 5 years to realize that they were perusing insights. The field of scientific visualization discovered that they needed a wide range of tools to reach the possible insights and that there was not one, great, end-all / be-all tool to give them that.

Ruth said this was not just a question of investigation, but it was also a way of expressing her ideas, her imagination. It was storytelling.

Lon offered that perhaps we're going the wrong direction towards realism. Char Davies came up with an abstract / surreal storytelling apparatus that had great emotional impact.

Paul thought that there was a danger of interpreting the past through our own memories of history, very real stuff (dirt, heat, hunger). One aspect of this is to try to deliver raw source material and also show the results of imagination.

Lon went on to consider the dreamy, the inaccurate. If something is imaged in not a fully accurate fashion, it left the mind open for multiple interpretations (sketches versus photoreal images). This might be critical, especially in areas of only estimated perspectives.

Ruth noted that the ancients didn't see the pristine conditions that we're presenting and react to them. How we see, we don't see with incredible clarity - how do we remember?

Rebecca reminded us that we should remember that when we remove an artifact from the ground, it becomes subjective to interpretation because of its loss of context. We must remember to keep it in context.

Alyn offered that we wanted to speak for inaccuracy and risk aversion. He had just finished a book on Fermat's Last Theorem. Mathematics is a result of false and true proofs. Tool builds needed to start, even if they would be proven wrong later. We shouldn't be risk adverse - we're learning along the way. Paul agreed. When he writes papers, he compares himself to the past - without it, he has so much more to do because he has nothing to stand upon.

Michael was interested in capturing the process in the field - helping the archaeologist to capture the experience. Kevin also added that capturing ambiance was important. Capturing the temporal aspect is important because there is a temporal context, an insight into the mindset. We need to allow the human perceptual system to catch up with the processing of all of this information.

Conclusions / Issues:

• What are the grand challenges for analysis?
• What should we be proposing as research projects for the future?