Project Jigsaw:
The Problems of Accuracy & Detailing in VRML

  • A VRML presentation models the real world by breaking it down into a series of lines in space. The more complicated the objects modelled, the more storage they need, and the more memory is required to display them;
  • Simple objects - such as a cardboard box - display quickly and easily; but complicated shapes - such as a complicated cornice, or the Euffel Tower - will take so much memory that most computers will be unable to display them;
  • Whilst we can envisage a time when displaying and manipulating such models will be trivial, this is not the case at the moment. The strategy we have adopted is to model Borobudur in as sophisticated a manner as possible, using AutoCad, and hope that, sometime in the future, its full glory can be handled in VRML;
  • For the time being, however, compromises have to be made, and these have involved: (1) the simplification of profiles; (2) the copying of one model onto many where possible (e.g. Buddha niches; the 72 stupas on the circular terrace); and most noticeably (3) the disaggregation of the stupa into constituant parts so that each may easily be loaded into a VRML browser: because of size problems, each gallery has been quartered, and each quarter is again divided into two section comprising (a) the main wall; and (b) the balustrade wall - that is, 32 separate VRML presentations for the four galleries, plus another four for the Hidden Basement;
  • In our presentations, the VRML vectors make polygons, and each poygon is "clothed" with a photographic image, in an attempt to give the user the "photorealism" of standard photographs together with the manipulability of VRML vectors. Of course, the dimensions of the photographs used impact heavily on the flexibility of the model; so once again we have adopted a strategy which will be robust enough for the future. That is, whilst our photographs of Borobudur are 1.6 megapixels each (about A3), we are currently using a cut-down series of images for testing purposes. We should be able to present the iweb-based user with a choice of large or small image-suites - always depending on network speed, of course;
  • On a fast machine with a good graphics card (Pentium Pro 400 with 128Mb RAM and a Diamond FFireGL 1000Pro with 8Mb RAM) the manipulation is smooth and easy. With a "normal" machine (Pentium 133 with 64Mb RAM and a common-or-garden video card) the presentation will be slow and jerky;
  • The presentation will also be offered on CDROM (very probably on at least three CDROMs), in which case the user should be able to choose large, medium or small image-suites. As technology gets faster, so our presentations should look even better.