A DPF WebApp prototype

A WebVR prototype based on DPF framework has been deployed in order to test selected dataset from San Marino project. Several features were added to DPF.js library, for instance better a support for mobile device orientation, better semantic encoding, and built-in support for positional 6-DOF controllers such as Oculus Touch for locomotion among the DPF network, right inside the browser. Here is a quick video of recorded online browser session:

You can test out sample demo with semantic interrogation for mobile here, and a demo for desktop browsers with DPF-Network here.



(Re)Building Forum Augusti


This is a first update regarding progress (july-august) on game-oriented assets production for the new Forum Augusti employed to create reusable, flexible and modular 3D components. The FBX models target utilization mostly in Unreal Engine 4 and SONY PhyreEngine (Playstation VR) – within REVEAL Project. Overall process started out from basic props and components (e.g. column, doors, braziers, etc…) most of them re-optimized to fullfil proper guidelines, including PBR requirements within VR fruition.


Most of per-asset workflow focused on geometry optimizations, level of detail, uv-maps and other aspects, taking into account a few considerations regarding VR best practices and in-game fruition. At this stage, produced FBX were imported in Unreal Engine in order to verify modularization, parametrization and procedural generation, described in a previous post. Such approach guarantees great flexibility for level designers or, if different hypotheses are validated during the process.

The video shows a short overview of a few experimentations right before the summer break, using produced modular assets, their parametrization and a first draft for PBR-based materials. Such tasks were also carried out in order to verify performance scaling at macroscopic level. More technical updates to come… Stay tuned.

Building Procedural Instancing Tools for Unreal Engine

When dealing with repeating 3D elements, a meaningful approach is to procedurally describe and generate them through a template and a transformation-list. The new upcoming version of ATON 2.0 for instance (yes, pun intended) already provides built-in procedural methods in order to instantiate several 3D objects by streaming a template (e.g. a column) and a transformation-list, thus greatly reducing bandwidth. On the desktop side, a porting of such functionalities from OpenSceneGraph to Unreal Engine 4 (ver. 4.16.1) began a week ago.

A C++ ActorComponent has been developed in order to read common ASCII files representing TransformList (position, rotation and scale) and a StaticMesh representing the Template. Such functionality can be employed into a compact and very basic Blueprint node (see figure) offering a flexible and manageable tool for designers, while maintaining all parsing and processing routines at C++ level. A TransformList file can be externally updated (e.g. generated by another software) and the procedural generation automatically rebuilds right inside Unreal Editor.

A second Instancer tool has been developed (this time, fully Blueprint) to offer a customizable tool to generate randomized instances (always in Editor) in a given radius with draping, overlapping options, scale variance and much more. The following shows an Instancer with trees (top left), another Instancer with boulders (bottom left), and a third Instancer (ferns) testing adaptive location on top of world geometries (right).


Adding another Instancer for grass, and tweaking of scale variance inside Unreal Editor.


There are more upcoming features… Stay tuned!

VR in your browser

During the last months several updates were rolled out for Aton – right before summer vacation. The WebVR one allows users to toggle immersive VR fruition of 3D models and archaeological areas through HMDs like Oculus Rift, HTC Vive and many others in WebVR-enabled browsers.

To get started with WebVR, you can download recent Chromium builds here and start exploring published 3D scenes through Aton Front-End. Among the goals of such feature is also realizing the so-called WebVR responsive concept, where the service adapts to your viewing environment by using fluid layouts, similar to Aton responsiveness for mobile devices (HTML5 UI, multi-touch controls, simplified shaders, etc..). Next steps will focus on proper traveling handling in order to offer a smooth HMD experience.


The WebVR update for Aton also couples with another upcoming update for the desktop application (ovrWalker), targeting performance for complex, multi-resolution 3D scenes and 3DUI research applied to CH – watch the video here. VR features in Aton will be further enriched, as well as massive upcoming updates related to node loading – so stay tuned!

Fictional 3D Landscapes: the Game of Thrones map experiment


This is an ongoing experiment aiming to create a fictional 3D map of the famous Game of Thrones TV show. Landscape Services were used to generate multi-resolution base terrain combined with multi-layer imagery to exploit Aton‘ simplified PBR rendering.

Click on the following image to launch the interactive exploration (WARNING: 3D map contains spoilers of previous seasons).

Click on this image to explore the interactive GoT fictional 3D map

Starting from a base imagery map, a DEM was created from scratch (isolating water, rivers, hills, etc…): in this case no geo-referencing (through world files) is involved, since world scale and extension are… quite fictional and pretty vague. Separate scene-graphs were built to handle simplified trees (mimicking those shown in opening credits), taking advantage of instancing capabilities, as well as a water level with own multi-texturing – all composed into a global scene. This experiment also introduces the new annotation features offered by Aton, to map major locations and/or events (for now). The scene is fully evolving: digital elevation model needs some fine-tuning, new annotations and 3D models of major cities / places will be eventually added to enrich the overall composition.


More info coming soon: this post will be updated!

A new PBR Model


Another major update for Aton is about to be deployed. A lot of work has been carried out to provide a modern, efficient and real-time PBR model. A lot of inspiration comes from Unreal Engine 4 (UE4 for short) and its advanced PBR system. WebGL world of course faces several limitations that need to be addressed, sometimes in “smart” ways or using approximation techniques (special PRT and SH solutions and much more) to reduce GPU workload.

A few samples using the new Aton PBR model.

Check out this demo, or this one.
The new, upcoming PBR system combined with RGBE model for Aton, supports now:

  • Base map (diffuse or albedo)
  • Ambient occlusion map
  • Normal map
  • Roughness map
  • Metallic map
  • Emissive map
  • Fresnel map
The new Aton PBR with the new real-time rendering engine. Textures modified from http://hrp.duke4.net/
Close-up on variable roughness, reflections and indirect light contribution (bottom cube faces)

The new PBR maps workflow also has the objective to be as close as possibile to a workflow involving UE4 (or other modern real-time PBR engines), to fully reuse such maps (e.g. “Roughness“, “Metallic” pins in material blueprint in UE4). Nevertheless, the new model is also compatible with “basic” workflow, such as the classic diffuse-only 3D modeling (or diffuse + separate AO, etc..). Screenshots below show sample workflow in UE4 using same identical PBR maps applied to cube datasets:

Sample workflow in UE4 material BP using same maps.

Of course these improvements will also extend to multi-resolution datasets (e.g. ARIADNE Landscape Services) to produce aestetically pleasing 3D landscapes by providing additional maps in input section, still maintaining efficiency of underneath paged multi-resolution. Furthermore, the Aton PBR system is also VR-ready, providing a realistic and consistent rendering of layered materials also on HMDs, on WebVR enabled browsers.


To import and ingest 3D assets generated using common 3D formats (.obj, .3ds, ….and much more) into Aton system and its PBR system, the Atonizer service will be soon available.

Stay tuned!

Useful Links

Light Probing system is here


A new massive update for WebGL Aton FrontEnd has been deployed.
The real-time lighting component – including GLSL vertex and fragment shaders – has been completely rewritten to feature a full light probing system, with advanced per-fragment effects.

The FrontEnd is now able to manage ambient occlusion mapsnormal-maps, specular-maps, fog and other information to correctly render different materials and lighting effects using a modern, per-fragment approach. Note LPs also contain an emission map, to consistently define light sources and also their shapes. Consistent rendering of normal maps required special attention regarding efficient computation with GLSL fragment derivatives (this will be explained in a separate post).

Chrysippus Head with LP: left without normal maps, right with normal maps. 3D Model created by E. Demetrescu

You can interactively change lighting orientation by holding ALT key and moving the mouse around. Try it live with:

Left with LP disabled, right with LP enabled. Click to open live demo and play with parameters
Light Probing applied to multi-resolution terrain datasets. 3D terrains generated by ARIADNE Landscape Services

These new features also apply to terrain datasets and, generally speaking, to all multi-resolution 3D assets published through Aton… Have a look at this landscape generated using ARIADNE Landscape Service. Yes, that means new options will be added soon to terrain service for processing multiple maps (i.e. specular maps for lakes, rivers, etc.) for 3D terrain datasets published online.

An Earth model with diffuse, normal and specular maps with unreal stars

More info and details will soon be published. Stay tuned.

Official project page for WebGL Front-End now live


The official page for WebGL Front-End – codename Aton – is now live. You will find a complete description of all its latest capabilities, demos and other showcase.

A live demo has been published about the 3D interactive Massenzio Mausoleum along with metadata presentation as well – including a brief description, reference paper, a few POVs and also a youtube video.

Improvements and new dialogs for WebGL Front-End


A massive update for WebGL Front-End has been deployed. Here’s a list of improvements and new features

  • Improved spherical panoramas handling and support
  • Added double-click hotspot feature: user can now double-click any point of the 3D model to smoothly focus camera on target. On mobile tablets and smartphones the same feature is bound to double-tap – try it live!. This feature is very useful on large-scale datasets to adaptively scale and focus targets or specific items
  • ccamImproved POV (Point-of-View) dialog: now the button shows an drop-down overlay with both permalink (3D scene + POV) and the bbCode – useful for text descriptions. During the exploration of the 3D scene the user can now simply press the POV icon (or press ‘c’) to show the drop-down dialog with current POV information and links ready for online sharing
  • ii-embedNew embed dialog (similar to POV dialog): it provides HTML code for use in your external web pages, enriching content with an interactive canvas
  • Performance boost in several areas
  • CSS improvements

Customizing the WebGL Front-End


This is a bit more technical post about customizing the main PHP component, responsible for main HTML5/js Front-End presentation and deployment on multiple devices. Here is a brief list of features and capabilities offered so far as GET arguments, useful for permalinks and/or embed in external pages.

“ml”: define nodes (3D assets) to be loaded from the online collection. For instance:

  • ml=ref.osgjs
    will load the single node “ref.osgjs” (basic coordinate system) – see here a live demo
  • ml=mytest/mymodel.osgjs
    will load the node “mymodel.osgjs” inside the subfolder “mytest”

Multiple nodes can be grouped by “;” – Examples:

  • ml=mytest/modelA.osgjs;mytest/modelB.osgjs
    will load both modelA and modelB
  • ml=mytest/
    is a shorthand to load an entire folder and it will load all nodes contained in “mytest” sub-folder. Note the “/” at the end

“pano”: loads a panoramic (equirectangular) image and attach to on-the-fly spherical geometry. Examples:

  • pano=mypano.jpg
    will load the equirectangular image “mypano.jpg” and attach to special sphere geometry – live demo

Notice you can have three different scenarios for interactive exploration:

  1. 3D models only (“ml” provided but no “pano” option provided)
  2. Spherical panorama only (only “pano” argument provided)
  3. Both 3D models + spherical panorama (“pano” and “ml” both provided)

“alpha”: enables a transparent background. This is useful when embedding the component in external web pages for styling purposes. Here is a demo with “ref.osgjs”viewer.php?ml=ref.osgjs&alpha – Notice how the background is now white. Try embedding it in another web page with its own background to see the blending.

“pov”: tells the Front-End to load a starting POV (Point-of-View) by using smooth camera transition. The POV string is composed by 6 values encapsulating eye (x, y, z) and target (tx, ty, tz) positions. Examples:

  • pov=-1.674 -2.996 1.159 -1.733 2.900 0.121
    will first load assets (a panoramic image and/or 3D models) and then load the POV – live demo

The Front-End of course provides shortcuts to provide user with current POV, thus allowing to share the current 3D scene and a specific camera position