Editorial note: This article contributes to the discussion on the merits and drawbacks of digital visualisation of cultural heritage, initiated in the September 2006 issue of the 3DVisA Bulletin (cf. Will computer modelling get better?, Review of the Phimai Model). Here, Michael Greenhalgh returns to the issues of veracity and accuracy of computer models of architecture, refuting the artistic, theoretical and practical arguments proposed by Daniela Sirbu in the article, Beyond Photorealism in 3D Computer Applications, published in September 2007.
COMPUTER NON-REALITY: FOR TRUE BELIEVERS ONLY!
Micheal Greenhalgh's riposte to Daniela Sirbu
I must confess to being surprised at Daniela Sirbu's introduction of various anaemic varieties of theory into the ongoing discussion of the veracity of computer reconstructions. When I hear the word 'culture', as Stalin nearly said, I reach for my dictionary. Some of us may need to do likewise to understand what point Sirbu is making. For what on earth have trendy philosophers to do with the matter under discussion which (to refresh the memory of our attentive audience) concerns the continuing lack of accuracy and detail in reconstructions of large-scale objects and settings made by computer? Shifting the argument to theory (thereby setting up a series of questions that the writer can conveniently address) shows the sleight-of-hand of the three-card trick, and avoids the issue rather than addresses it. This shift is clearly seen in the question, 'why focus on photorealism when no true photo camera is involved?'
That is, Sirbu chops logic, and tries to suggest that some element of remediation (whatever that is) is involved in computer graphics, strongly suggesting thereby that the computer corrects and updates the photograph. But she doesn't chop the logic well, for there is little difference result-wise between light falling on an emulsion or on a CCD, since the results of both technologies can be measured in terms of resolution, and it is the normal tactic of the latest snake-oil salesman to disparage the competition. So where on earth does remediation come in, except in Bolter and Grusin’s desire to egg the pudding of computer graphics with a theory which has been welcomed with open arms by the electronic community? After all, in the academic world you’re no-one today without a theory to pile up against Deleuze, Heidegger, Lacan, Baudrillard and the rest – although it is noticeable that true computer scientists manage to get by just with ideas and code. So why side-step the real issue by recourse to theory? Or, to cite another pseudo-philosophical question, Why are you Saussure about things you know Foucault about? Indeed, following Daniela Sirbu's recourse to theory to give us all a lesson in the nature of reality, we should perhaps be grateful that she did not immediately fasten on the notion of truth as well. What is truth? said jesting Pilate - but then he was too worldly-wise to stay for an answer.
My original assertion bears repeating, namely that no computer reconstruction yet made reaches the level of accuracy of actual photographs, nor will one ever do so, because the process involved is indeed a reconstruction of elements which simulate the real world and do not reproduce it. None of the examples suggested by Daniela Sirbu is remotely like the real world. I maintain that the only area in which such reconstructions have a role to play in academia is where the object or setting no longer exists, and is indeed being reconstructed from suggested rather than necessarily proven elements - as with buildings on an archaeological site of which only the ground-plan and other vague comparanda exist. But to use a computer to take an actual setting to bits, and then rebuild it tediously in the computer, seems to me a waste of time – although to Sirbu 3D computer applications go beyond photorealism, although she never explains how, let alone why.
My objection to any recourse to theory is that it does not bolster arguments about computer accuracy, but simply side-steps them. Read the literature on computer modelling, and you will see many practitioners happy to aver that the world the computer constructs is not the real world, but a parallel one. This is not, please note, because of any horror of the world in which we live (and in which we find the objects, sites and settings we study), but because excuses are needed for a technology which is incapable of reproducing it without extravagant amounts of time, money and skill in varying measure.
Computer gaming shows the way, as it has provided the motor for the development of several innovative technologies to speed up the creation of worlds. We should enjoy these for what they are, but stop pretending that constructing such worlds is any substitute for academic research focused on the real world. Such new worlds are certainly creative (cf. Sirbu’s artist behind the technology), but have nothing to do with the real one, as can be seen through the texturing, which continues to be anaemic and soapy. Ten years ago I too was a true believer - seeing the possibilities of VRML, and looking forward to its further development in what I felt sure was a field as fast moving as various others (memory, image processing) in computers. But little progress has occurred, except in the automation of data-gathering.
Please note that I do not argue above that accurate reconstructions of monuments cannot be generated in a computer - only that to do so requires extravagant amounts of time, money and skill in varying measure. For example, 3D scanning has been used to model the Great Mosque at Sana'a in Yemen (7th to 12th century). The French and Yemeni team used Trimble 5600 and Trimble 3600 Total Stations (costing in the tens of thousands of dollars) for the control survey, with the set measurement interval for the scanning being 15mm - and this for a building of 2600 square metres. The team note that ‘In contrast to photogrammetry
techniques, the scanner’s acquisition
time is proportional to the
detail desired.'; and ’ENSG students [i.e. the French Ecole Nationale des Sciences Géographiques] from various classes have contributed towards perfecting the best methodology, but there are still several hundred photogrammetric and point cloud pairs to be combined.'1 The interesting write-up of this labour- and technology-intensive project (cited here) has five illustrations, four of which are photographs, and only one of which is a computer reconstruction.
The bottom line is that computer mapping and reconstruction of real-world objects and spaces do not and will never look like the real thing, because of a continuing inability to deal with detail and accuracy, let alone the crucial matter of texturing. In the reconstructions of the Teatro Olimpico in Sirbu’s piece, why would any scholar wish to use them, rather than viewing photographs? After all, they are but approximations, and are much less accurate. So let us be serious about the extent to which such reconstructions can and should be used in research. After all, if I want to know about Gaul, I read Julius Caesar – not Astérix.
1. N.N. (2007),'Scanning Yemen's Great Mosques', Technology&more, A Publication for Surveying and Mapping Professionals, Issue 2, pp. 2-4.
© Micheal Greenhalgh and 3DVisA, 2008.Back to contents