Still alot of work to do to finish this film and there’s some obvious flaws but I decided to post it after my Masters thesis submission nevertheless. Next post, (probably around graduation time), will be the final product!
Latest
Archive Simulator (ext. & int.)
The Archive Depot scene is finally complete after weeks of non-stop rendering. The image shows archive crates arriving at the Tadao Ando inspired ‘Ito House’ in Tokyo, wonder if he would approve of its re-use?? Rendering has now began on the interior scene, where the magic really happens, watch this space…
Two Left Feet
This really is the last of the model construction as time is certainly ticking! The diffuse and difficult to photograph surface texture to the boot, (yes there was really only one) meant a difficult build. The texture atlas is far from perfect but it shall do for now…
Setting the scene
The ‘Archive Exchange Depot’ from scene 2.
Following on from the narrative discussion in the ‘About’ page, the short film has been organised into 3 sections:
Scene 1_THE BEACH: Live action footage of real objects washing up on the shore with myself collecting and labelling the objects. Wider camera pans reveal cgi ‘archive crates’ scattered around the beach awaiting collection. Digicam effect recording display with ‘scan cam’ info/date/place.
Scene 2_THE ARCHIVE EXCHANGE DEPOT: (full cgi) Technology in the real world stands stagnant in the mechanistic functionality of old, maintained only to serve the needs and desires of the model simulation. Recieving constant deliveries from the ‘real world’, the Archives Exchange Depot recieves, sorts and dispatches the never ending stream of valuable archives to the designers and engineers of hyperreality. Animated archive crates move in x and y directions through the grid-like skyline watched by, this time, a 3d photoscan of myself.
Scene 3_ THE SIMULATOR: (cgi and live action) An interior, virtual reality facility for examining, analysing and interpreting simulated reality of digital artifacts and environments from real world archives. Storyline TBC……
Custom IBL lighting test
Tidal Points is a digital project dedicated to the representation of the real; real objects and a real place in the form of a virtal-installation. Following Christian Bloch’s excellent ‘The HDRi Handbook’ I embarked on the creation of my own 360 degree VR mirrorball for capturing accurate on-site lighting condtions that can be used to convincingly re-light digital models placed in still or video imagery.
Firstly, the mirrorball costruction. A cheap garden ornament purchased from ebay mounted to an equally cheap and nasty tripod through various nuts, bolts and drywall hangers (those years in construction were not entirely wasted!).
Next up. On site hdri photography. Numerous exposure values of the same image that will be merged in Photoshop into three 32 bit floating point, HDR Radiance files. To be safe, 9 pictures taken, a rotation around the mirrorball 90 degrees and 9 more, then 180 degress to finish.
In HDRShop the images are cropped and unwrapped through panoramic transformations and again saved as HDR Radiance files.
Back in Photoshop the 3 layers, offset and aligned, are masked and cloned to remove artifacts (the camera, tripod and photographer) as well as go some way to remove the ubiquitous pinch caused by flattening a 360 image. Saved again in HDR Radiance format and done. Its certainly not perfect, far from it but it will do the job..
Now over to 3DS Max for a quick test. 1st loaded as a spherical environment in the material editor and then instanced to a Background/Environment Switcher. Then a background image in a Matte/Shadow/Reflection slot which is also instanced to the Switcher. The Switcher is then instanced to the Environment channel. The addition of a Skylight (with ‘use scene environment’ checked)and a MR Area Spot for shadows. Job done.
Finally, the last of the models…
The half tyre model represents the last of the point genrated mesh models. Again, a diffcult model to get right and by no means is it perfect. The 20,000000 (million!) poly model emerged from around 65 photographs. A lack of texture detail and the black diffuse colour meant that the quality of photography was particularly important. Placing a bright, textured covering under the model ensured a number of fixed algorithmic points for Photofly as well as providing backlighting to the underside of the tyre, the downside is that texture relections must be removed in Photoshop prior to baking. In MeshLab the usual parametric cleanup and hole filling took place followed by the now, much loved, Poisson filter, (essentially the creation of a solid and useable piece of geometry), over for a quick ‘smooth’ in Mudbox before exporting to 3DS Max.
In Max the retopologised model (which took far too long) proved impossible to work with and so I decided to break it into 2 components, the tyre and the rope, allowing much more control over each element, i.e. in the normals and diffuse texture baking parameters. Overall a decent enough model yet the rope looks too solid and lacks the physical attributes of the real object. Perhaps some remedial modelling could rectify this however the clock is certainly ticking….
The lobster pot was always going to be last, after much time and effort on prior models I imagined I’d be in fine stead for tackling the intracies and difficulties of such an awkward piece of geometry. Im glad I did. To attempt to photomodel this beast would be a step too far and perhaps, madness. The removal of some or all of the netting would make the process feasable yet what is a lobster pot without it’s net? Instead I opted for the good old poly model option, is that cheating or just knowing your limits?
So, finally the modelling phase can make way for some clearer storyboarding, on-site filming, hdri photography, animating, rendering…………
more rubbish (models)
Plastic bottles consisting of 41% of the beach debris and metal cans taking up 14% of the survey material, are not part of the point cloud work. Having little detail or texture information of interest they were simply poly modelled and then deformed. Textures took on a ‘beached’ and ‘washed up’ appearance through Photoshop manipulation. The gas bottle and traffic cone, as on-site objects and only partial scans, were modelled through the Project Photofly process, the raw and untreated models then simply acted as accurate reference material for new, poly models. Partial, projected texture atlas’s were refined and completed in Photoshop.
Trainer shoe model
Project Photofly came up trumps with this shoe model. A top and bottom scan of around 70 photographs has created a very accurate 3d model. Exported to MeshLab for clean up, merging and remeshing with Poisson filter and then over to Mudbox for a quick topological cleanup. In 3DS Max the very low poly retopologised model took fairly kindly to projected normal mapping. Textures lost in Meshlab merging (apparently soon to be rectified) used 2 projected models to create uv’s which were then merged into 1 in Photoshop. For materials the map was split into leather and sand components (reflective and dirty matte with slight displacement), combined in a composite material slot. Perhaps this is the most accurate and believable yet?….
paint brush and railway sleeper models
66 photographs were necessary for the paint brush model and even that left a faily noisy and messy job. A hefty resculpt in Mudbox with some additional bristle indentations, shipped back to Photoscan for texture alignment and then to Max. A protocol Im getting used to now; low poly retopo with normals and texture atlas projection via the render to texture tool set. Next up, the railway sleeper, weighing in at over 7 feet long remains in situ (or perhaps now somewhere out to sea?) The eagerly awaited release of Autodesk’s ‘Project Photofly v2’ has arrived. Very user friendly and a very quick turnaround considering the density of points/mesh. Using only 16 photographs the sleeper consists of 2 identical halfs which is a story in itself….
Project Photofly, worth the hype? Yes. But it does have its limitations. No merging models or texture reatlasing or real editing capabilities.
tidal points 