Reference 3D Models In game development, reference objects are represented by a simple cube. These cubes can be used in the level creation process and are replaced by fully modeled objects. A simple feature to quickly create cubes of appropriate sizes was added to the plugin. However, in the context of architectural problems it is important how the surroundings it to a human.
The plugin allows a quick import of a correctly sized adult human reference igure. This enables the artist to verify that, for example, furniture used is scaled correctly. It is common practice to export assets and reuse them as size reference . The plugin scans a directory for existing models and provides a drop down menu to import the selected model. Figure 4. Primitive Collision Objects 3D assets in UDK have two parts, a model part that can be seen visually, and a collision model that is used for calculating collisions.
It is best practice to provide for each UDK model a speciically designed collision object. When no collision object is provided it would allow the UDK players to walk through the object.
Manually creating collision objects is a tedious task. The UDK Editor provides a tool to create collision models, however, this has to be done for each individual object. The plugin provides three ways to create basic collision objects: Box, sphere, and duplication. The box collision object can be used for most parts of a building, like windows, walls, loors, and ceilings. The plugin calculates a bounding box around the object and then creates it as a collision object.
The sphere collision object is suited for complex and round shapes, like chandeliers. The radius of the collision sphere is calculated based on the size of the bounding box. Collision objects created by duplication provide a quick way to test complex objects in the UDK. The plugin names the collision object correctly and parents it to the original object.
This ensures that object and collision object move together. The lighting information is baked into textures . The script automatically re-projects the UV coordinates, and by doing so, gives every polygon-face its own UV shell. Export Tools The most effort for the plugin was required for improving the export-import worklow. Besides exporting multiple objects it also supports the asset management ile structure. A method was implemented to export all positions of the object and import the positions into UDK. Thus the user has to manually import the exported assets. Exporting all iles in the same folder structure allows rapid re-import of assets.
The plugin has the option to always export FBX-iles to the same folder. This ensures a standardized format and uniqueness of the names avoiding common errors. Multiple Mesh Export The standard export operation exports all selected objects to a single ile. UDK, however, needs every object stored in its own ile. Initially we created a method to move the object to the origin, then export each object as a single ile, and then move the objects back to its original position.
While this is the correct way to export optimized assets for the UDK, it is only needed if assets have to be reusable. As in architectural projects the geometry of the building is unique, it makes this feature unnecessary. To reduce the complexity of the plugin, we removed this feature in the inal version. T3D - Level Export To export the level correctly, besides the model itself, the positions, rotation, and scale of the assets have to be stored, and then imported to a UDK map. While researching the problem, no solution could be found using the oficial documentation. Even the deprecated oficial ActorX plugin did not offer this feature.
However, we discovered that UnrealText ile format. The main use for the ile format is to allow the import of custom BSP-Brushes. This ile format is deprecated and no oficial documentation is available for the ile format . To investigate the ile format, we exported a level and discovered that information is stored in plain text. The text is structured simililar to a markup-language. Methodically we removed data structures and then reimported the ile to see if the remaining information was still importable.
The initial ile had lines. We noticed that texture iles were represented in binary. Simply removing the binary data from the ile resulted in errors. We removed all information not directly related to the static mesh UDK actor. The import of the remaining 24 lines of code worked without problems. The ile now contained just information about attributes regarding the actors that were being set by default.
As we were only interested in the location, rotation, and scale attributes of the object, all other attributes were removed. The ile was reimported successfully. However, there were no rotation and scale attributes deined in the ile. We speculated by using a different actor and setting values for the rotation and scale attributes that the information would be stored in the t3d ile.
An example of a t3d ile is shown in the appendix. It is important to note that this method only works for objects that are already present in UDK packages. In addition it exports the t3d ile. Due to the lack of a direct connection between Maya and UDK the user must manually import the FBX-iles to the correct package, then save the package, and inally import the t3d ile. The PyQT interface was not used as no callback functions were required for the interface. Visual cues from the Maya Attribute Editor interface were used to design the plugin. Limitations of the Plugin During the import-export process the user has to perform a short series of operations.
If the user forgets to save the package before importing the t3d ile, then the ile is not imported and no error is displayed. The plugin does not provide a solution when using translucent materials. For these materials the settings have to be corrected manually. The automatic lightmap generation works only when used for technical objects. Errors occur when used for organic objects. The plugin eliminates several parts of the FBX worklow see Figure 4.
In fact, the worklow is now almost identical to any 3D asset creation worklow, the only addition is to use the plugin to export the scene and to import it to UDK. First we created a scene with 10 basic spheres positoned at random. Then we transferred all objects with collision model and lightmap, and positioned all objects correctly in UDK as quickly as possible. Using the FBX worklow the time required to complete all operations was 7min 10sec.
When using the UDKToolbox plugin, all operations required 40sec. UDK Framework In this chapter we will discuss how we extended the Unreal Engine to serve as a framework for architectural visualization. UDK Framework Design Our goal was to develop an open framework that not only could support our prototype for architectural visualization but to also allow rapid display of other types of architecture and to make it extensible for including new features.
To create a UDK Game, four major components are needed : A game deinition written in UnrealScript , an Unreal Package containing the assets , an Unreal Map deining the environment , and the game engine coniguration iles. A standard text editor is used to create Unreal Script iles and coniguration iles. UDK Maps Maps also called levels store information about the environment, properties of actors position, rotation, etc. A UDK actor is deined by a reference to an object in an UnrealPackage or an instance of an actor-class- object.
Additional maps can be loaded during runtime using level streaming methods. They are optimized to store and manage assets. The package compresses and optimizes the assets to minimize memory consumption. The size of a package should not exceed MB. A special package component is a Prefab.
Prefabs are collections of actors and associated Kismet that are stored in a package and are reusable. For example, to open a door in UDK, the mesh, a trigger, a matinee sequence to animate the door opening and its Kismet script are stored as prefab. UDK has several types of games available. This game type can be modiied to suit architectural visualization.
While visualization is not a game, to access the rendering capabilities of the Unreal Engine, a simple UDK Gametype without game elements must be deined. To accomplish this goal, we identiied potential elements that are present in every type of architectural visualization. If possible, these elements should only be created once and then reused in further architectural visualization projects.
The feature is used to create a seemingly endless environment without the need for loading screens. In our framework, however, we are using it simply as a layer organization tool. Multiple levels are used to manage the featueres in dedicated maps. In addition, the level setup can be easily extended with features if necessary. UDK Packages The uniqueness of every architectural project is the actual building. Separating the geometry of the building and the materials into its own package allows the materials created for the visualization to be reused.
Unreal Script Classes The framework includes an architectural visualization game type class to control the engine that does not have to be further modiied. However, the coniguration settings can be set to a low-end environment removing the need to modify these iles further. Asset Management UDK organizes its assets in packages. For optimal performance, the texture iles have a length represented by a power of two but not larger than To use the Flash elements they have to be imported into their own Unreal package and then applied as a texture to an actor.
Materials Materials used for architectural visualization should support the following attributes: diffuse relections, specular relections, and refractions. To demonstrate the material shader capabilites we are using a Utah teapot as standard reference object. During the material creation process the artist should be communicating with the architect. It is easy to create material shaders that are visually pleasing, however, it may not be feasible to build the building with the suggested materials . It has a diffuse and specular highlight shading component. Diffuse materials scatter light in every direction.
In our surroundings most objects have a diffuse relection component, for example, plaster walls. Figure 5. Relective Materials The basic material shader does not directly support specular relections due to performance optimizations. A method to create specular relections requires the use of a lightprobe to sample the environment.
The sampled data is stored and then used as texture map for the relection. The node stores the relection data as cubemap texture; this texture has to be projected onto the speciic object. The converted texture is then used as diffuse relection. For other materials that have a diffuse and a specular component, the shading network has to be extended to allow the blending of the diffuse property with the specular property.
In the following example, see Figure 5. When dynamic is enabled, UDK constantly updates the material, which can lead to performance issues.
Visual Data Analysis | Zuse Institute Berlin (ZIB)
Static relections are calculated only once when the level is generated and then never updated. In addition to the potential performance issues, the method requires that for each object that uses a direct relection component, a unique material and accompanying light probe must be placed. This process can be very time consuming for the artist. It is generally discouraged to use relective materials in UDK. Refractive Materials Transparent objects, like glass, cause the light being refracted by the material.
The model has additional attributes to modify the opacity and distortion. When looking at a rounded transparent object, like a glass, the object is usually very transparent in the center; however, on the edges of the glass it appears to be less transparent. The light is refracted and relected simultaneously; this effect is called Fresnel relection. This node is added to the shading network to control the opacity.
A similar setup can be used to control the distortion factor. A major issue when using this material is that it has no option to enable translucent shadows as well as stencil shadows. Refraction index values cannot be directly applied and distortion attributes have to be manually adjusted. Glowing Materials Lamps when turned on emit light and have a visual glow around them.
The basic material has an emissive attribute to make objects glow. Values above 1. However, the property can be used to create the material of the lampshades and by placing light sources near the lamp it appears that the lamp would be actively emitting light into the scene. These nodes function as named variables. This allows complex shading networks to be easily reused for multiple variations of the material. In addition, the system has the possibility to modify these values dynamically. Light Placement When lighting for animation is used it is common practice to use 3-point lighting: a key- light, ill-light, and rim-light.
The light sources in most cases cannot be seen and can be simulated by using basic virtual lights . However, this type of lighting setup is created based on the camera position. For lighting in interactive 3D environments it is important to place lights in a plausible fashion and independent from the camera. In addition, because the user can explore the environment, in most cases the user should be able to see the source of the light. The process requires all levels to be loaded. All lights visible are used in the process regardless of its level location.
For each individual object the UV coordinates for the lightmap are used to generate a unique lightmap texture. This texture is then loaded at runtime. Dynamic lights are constantly calculated during runtime; they do not have any effect on the lightmass system. Due to performance issues Epic Games recommends having at most 3 dynamic lights in any given scene. Indirect Light Lightmass Lightmass is a static global illumination system. As with static lights, the indirect lighting information is stored in the texture of the objects.
The lightmass process is very complex and drastically increases light calculation. It is important to note that when using toggleable lights only the irst state of the light is used to calculate lightmass effect. During runtime of the game increasing or decreasing the light intensity has no effect on lightmass. It is for the same reason that dynamic lights do not inluence the lightmass effect as it has been statically calculated. Photometric Data Especially in the ield of architectural visualization the photometric data is used to imitate realistic lighting.
The default maps of UDK have a skydome object and a directional light. The material is set up in a way that the direction of the directional light creates a specular highlight that can be perceived as a sun . This setup is limited to a speciic time of day e. Basic Sun cycle We started to implement a basic sun cycle that rotates a sun sun disk and shadows around the building and shifts the color of the sky depending on the time.
This material was constructed so that the sky color can be controlled and at the same time a lighting channel paired with the directional light visualizes the sun disk. To enable the material to be modiied over time, an accommodating material instance was created and applied to the skydome. In Kismet a Matinee node was created. An entire day cycle was animated by setting key frames for rotation and material properties.
Shadow Artifacts We tested the system using a simple test environment containing a loor and a box. The shadow quality was too low to be recognized as shadows. The shadows were rendered in multiple stripes. These stripes also appeared on surfaces that should not receive shadows. When animating the time of day system, the stripes started to move and change forms.
Fixed time model To avoid the shadow artifact issues, a ixed time model was persued with four times of day being implemented: Morning, Midday, Afternoon, Night Initial Level Streaming Setup Our initial concept used the level streaming technique to switch between the different basic maps as provided by the standard UDK for the four different time of day settings. To ensure that the correct time of day is displayed it must unload all other time of day maps.
UDK does not compute multiple DominantDirectonal lights correctly. This behavior is not documented, as levels usually only have one DominantDirectional lightsource as sun. It was not possible to directly adjust the color of the light to relect the time of day. Instead we used a workaround by creating a Matinee sequence with the duration of 0 seconds to set the lightcolor attribute. Final Implementation By using the level streaming method it is tedious to extend the system to support additional day times, also the feature requires multiple map-iles. We further analyzed the different elements contained in the default maps.
We discovered that every map had the same basic skydome element, however, the material was different depending on the time of day. This setup eliminates the need to stream levels and, at the same time, makes it simpler to extend this feature. Exchangeable Environments Exchangeable environments demonstrate the lexibility of the visualization capabilites. As proof of concept, the user is able to experience the building in an urban environment, as well as in a nature environment. Nature Environment While most of the population lives in cities, it is a common marketing practice to surround the architecture in a nature environment to make the building look more appealing.
Nature environments have two major components, the landscape hills, lakes etc. The system interactively modiies the heightmap and displays the result. The landscape actor can be further modiied to include transparent sections. This feature is useful when the viewer should be able to access basement sections of a building. Due to performance optimizations only heightmaps with speciic dimensions are supported. The minimum size is x, the maximum size is x, which is equivalent to To support such complex geometry, the UDK Landscape system automatically reduces the complexity of the geometry based on the view distance.
The shading system provides the possibility to blend various textures based on the heightmap information. For example, in a mountain environment the highest points could be covered in snow, while the rest of the mountain is covered in grass. We created a x height-map to create the surrounding landscape and stored it as its own level. The UDK Terrain system is not compatible with mobile devices. This program generates trees based on node deinitions. The various nodes can be given random values to create variation of the tree models. The user can create a brush made up of multiple objects and the density, rotation, and distribution can be deined by a randomizing function to create the appearance of a natural distribution of the objects.
We used two types of trees available in the default packages of UDK and painted them onto the landscape actor. This type of effect can be created by using the UDK Particle system. This would allow time of year visualization. We did not explore this further. Urban Environment In urban environments the shadows of surrounding building greatly inluence the lighting conditions of a building.
Our system should visualize these effects. The viewer, however, should not be distracted by these buildings. Skyscraper We created a grid of eight skyscraper sized buildings that surround the property. These skyscrapers would completely block the sun so that minimal sunlight would reach the property. The shadow would not be visible on buildings far away. To eliminate this effect, we increased the maximum shadow draw distance, even though this has an effect on performance.
Cityscape Visualization Data As alternative to the surrounding skyscrapers, we tried to gain access to real world cityscape visualization data. We discovered a method to extract data from the program Google Earth 3D . The method uses the program 3D Ripper DX. Due to potential legal issues we did not further explore this possibility .
Environment Switch To switch between the cityscape and nature enviroment we used level streaming. The viewer starts falling into ininity. This issue has been resolved through introducing player restrictions in the Helper Layer. Helper Layer The helper layer ensures player movement restrictions, correct level streaming, camera setup, and basic lightmass optimization. Level Streaming This layer has a basic Kismet setup that initializes the other features of the visualization. This ensures that all levels are loaded. The ield of view angle can vary depending on the visualization.
We modiiy the value using a Kismet console command node. However, the user needs to be restricted to only move within the boundries of the created environment. Alternatively, invisible collision objects can be used to restrict the movement. We used the loor as an invisible barrier to prevent the player freefalling when the environment is switched. The fence is set up so that the user is restricted in all levels in the same way. It is best practice to give the user some indication how far he can go. We used a volume that is based on the size of the player barriers.
Visualization Menu Menus are built in Flash and interpreted in Scaleform. We require menus for time of day settings and to swap the environments, as well as for future features. Scaleform interprets Adobe Flash iles and renders them into the 3D environment. The command was originally designed to be used to communicate between the Flash Player and the Web browser or ActiveX component thus limiting the command to broadcast strings.
Two buttons are available for the environment control. Mouse cursor The user should be able to use the menu using the mouse; however, UDK automatically hides the mouse cursor. To enable the user to see a mouse pointer an image of a cursor was added as the top layer of the Flash ile. An Actionscript command enables the mouse to drag the image, thus visualing the mouse position. All buttons are stored in an array for extensibility of the menu design. When a new button is added to the GUI, only the array has to be updated, no other code has to be modiied.
Fscommand event nodes are added to control the environment and time of day. UDK Coniguration We attempted to keep the core UDK coniguration in its default state to enable rapid development of visualizations without the need to modiy code. Originally we started working on a visualization gametype that disables all game features of the game engine. This gametype allows the game engine to be used without any game rules, which is ideal for visualization. Architectural Visualization This chapter covers the creation of a architectural visualsation prototype.
No CAD-designed houses were available free of licencing fees; bitmap images of blueprints, though, are available online. The bitmap blueprints were lacking scale information. Using the reference photos of the house, the size was extracted by measuring one of the more prominent windows assuming a standardized window size. It is common practice of modeling houses via the block extrusion technique .
A polygon cube is created and the polygon faces are extruded. The downside of this approach is that the model is complex and the associated collision model is also complex. This can cause performance issues and faulty collision models in the Unreal Engine. We decided to use a modular modeling technique . For this type of modeling the reference images and the blueprint have to be visually analyzed to identify repetitive components, like multiple large windows.
These components are modeled as unique components. The major beneit of this technique is that the components used only need to be modeled once and then can be reused. This reduces the overall time needed for modeling. Other beneits of this approach are that if errors occur in the polygon topology they are easier to ix. In addition, it is possible to swap modules that are similar size, making it easy to create variations of the model. UDK also beneits using this method, because most modules created have a shape similar to a cube, making it possible to use primitive cubes as collision model.
Figure 6. Interior Lighting Models of lamps were created and the GUI was extended to support interactions with those lamps. Lamp Models We created three different types of lamps, a large ball shaped lamp, a lat lamp, and a chandelier. The design of each of lamps followed the same principle. A mesh for the entire lamp was created. An alternative process to have similar control of the material would be using texture maps, however, for each lamp a unique texture map would have had to be created and the material network for each lamp recreated.
The lamp model has no collision model due to the fact that lamps are hanging from the ceiling and usually do not obstruct walking in the room. In addition, several controls were added to display light combinations. The buttons controlling the lights were organized in rows according to loor. Kismet Script When toggling the lamp the light and the material have to turn off or on.
To be able to control the material effect individually, for each lamp a unique material instance of the glow material was created and applied to each lamp. The glow property of the material is controlled by a Matinee sequence. To simulate incandescent light bulbs a zero-second animation was used to toggle the glow property on or off.
These events have to connect to all individual light sequences. By disabling the counter and manually setting the index, the node can be used as an array node. The switch node is useful to preserve existing light combinations when more lights are added. Frame Rate We measured the performance of the system in three environments. User Study We conducted a qualitative user study to igure out if potential users would want to use such a system. The remainder of this text is structured as follows: In the irst part, the conditions, tasks, study design, and participants are described, in the second part, qualitative feedback, and observations are presented.
The last part has a discussion of the outcome. Conditions The study was conducted in a living room environment. The prototype was running on a desktop computer with following speciications: Intel i 3. The monitor used a full HD resolution x The prototype was reset and running before the participant was seated at the computer. Using the menu, the time of day could be adjusted, different light settings could be applied, and the environment could be changed.
The initial settings of the map were: All lights active, nighttime lighting, and the participant is facing the entrance of the building. Tasks The participants had to perform ive tasks: 1. Walk to the second loor and toggle the light 2. Switch the environment to the nature environment 4. Switch lighting scenarios 5. Activate the midday time of day setting 7. Study Design In the beginning the participant received instructions that he can move using WASD- Keys, orient himself using the mouse, open the menu with R, and activate lamps with E-Key when standing in front of lamps.
The participant was given the possibility to get familiar with the controls before attempting the tasks. When the participant indicated he was ready, he was given one task. After completion of the task the next task was given in numeric order. After completion of all tasks, the user was presented the survey.
While answering the survey the user was given the opportunity to use the prototype to clarify questions. The user was permitted to use the alternative input method using arrow keys; however, the input method was not explicitly shown to the participant. Participants Ten participants 6 female, 4 male, aged 25 to 65 volunteered for the study. The participants were home makers, media informatics students, and professionals from various ields, for example, medicine, chemical engineering, higher education, a tool and coniguration manager, and an IT consultant.
Only one participant was left handed. Five participants were experts using computers, three experienced, and two had no or little experience using computers. Only two participants were experienced using 3D authoring software, one was a Maya expert, and one an experienced AutoCAD user. Four participants had no prior experience playing video games, two participants had some experience, and four participants had very much experience. Games e. Qualitative Feedback After inishing the tasks, the participants illed in a questionnaire in which they ranked each question on a scale from 1 to 5.
Orientation and Menus Overall orientation created no problems for seven users. Two participants had dificulties using the mouse to look around. Only a single user had dificulties using the mouse and keyboard as navigation input. The majority of users had no problems entering the house. Two participants mistook the glass windows as doors and tried to enter. The majority of the users had no dificulties going to the second loor.
Three participants had dificulties walking up the stairs to the second loor. The major problems were that, while walking around the irst loor, due to missing railings they fell back to the ground level. Light Interaction It was observed that using the E-Key to control the light caused problems for all users; however, in the feedback ive participants stated that they had no dificulties.
The menu was well received. Six participants answered in the survey that it was easy to control the kitchen light with the menu. It is important to note that the menu gives no indication what button controls the kitchen light and all participants controlled the light by trial and error.
No users had problems switching between different lighting setups. However, four users were unclear why such a feature is needed. Three participants were interested if they could place their own lights or move the existing lights. Time of Day The entire feature set of the Time of Day possibilities was very well received.
Table of Contents
Users actively explored the different settings, even though it was not part of their task. Several participants liked how the shadows changed and were amazed of the large differences in length of the shadows. The predeined points in time were suficient for all participants. Midday, night, morning and afternoon settings were perceived as accurate. Nature and Cityscape Environment The ability to switch the surroundings was well received. One participant stated he would have liked even more environments to choose from, especially he would have liked to see the building in a typical urban environment.
Four participants liked both environments equally. The cityscape environment was preferred by three participants, whereas three participants preferred the nature environment. It is interesting to note that all three participants that preferred the nature environment were female and all three participants that preferred the city environment were male. For four participants the loading time required to switch between the environments disturbed their feeling of being immersed in the experience, however, for four participants the immersion into the experience was not disturbed.
Real Estate Marketing Viability All participants liked the visualization. They had no dificulties imagining how the building would look like if it would have been built in reality. Eight participants were interested to see the building in real. However, four participants stated they would need more information on the project. All participants were very excited by the prospect of having such a visualization embedded in real estate websites.
Almost all participants could not imagine to use such visualization on a mobile device. Two participants stated that the interactions with the hands on a touch display would occlude the display, thus making the visualization useless. One user was interested to use a mobile version of the visualization.
The participant stated that it would be enjoyable to brag about his house and letting people virtually walk through his house. Uses of the Visualization The users were asked if they would come up with potential use cases for the visualization. Communication Medium for Architects The biggest complaint of the participants of the study was that architects not always anticipate the needs of people living or working in the building.
They believed the visualization was an ideal tool to improve the communication between the architect and the client. She ended up with a room that had the right proportions but was half as big as she would have needed. By using the prototype she felt conident that she understood the correct spacial dimensions of the room. Another participant stated that in the hospital she works for, the architect planned a room encompassing an MRT machine. He was unaware that the machine had to be removed on a three year maintenance schedule.
Thus the building was not designed for this requirement. At the time, the head of the hospital was busy iguring out how to read technical drawings and was not fully focused on optimizing the building for the speciic needs. By using the visualization, the client could have fully focused on the building and the potential problems with the design; also the staff could have provided valuable input. Unlike architectural lythrough animations one has the possibility to freely explore the building.
The participants especially liked the beneit of being able to explore potential properties in their web browser and to eliminate the need to visit the sites. Interior decoration Two participants would have liked to have more features, like being able to place furniture into the visualization, enabling the visualization to be a tool for interior decoration.
Idea pitching tool for architects A participant suggested that the tool would be an ideal way to persuade investors to invest in large building projects. The increased interactivity and the immersion would be an ideal way to get investors interested in such projects. Learning tool One user suggested the prototype could be used to teach architecture. Beneits of the visualization The users listed the various perceived beneits of the visualization. Control of the Presentation Participants stated that they liked having control and time to think about the architecture.
The size of the building was clearly communicated. In addition, having control of the camera offers the possibility to view any angle without dificulties. This allows detailed views of the property, eliminating the possibility to hide an unfavorable chimney or crooked room. The users perceived the 3D representation superior to 2D technical drawings.
Especially people without training or the ability to imagine the architecture have a higher accessibility to architecture. Relocation and International Relocation The visualization saves the time to go and see the property. The ability to view the property independent from the current location would make the search for a suitable property much easier, especially in the context of international relocation.
Downside of the visualization The users listed the various perceived downsides of the visualization. Navigation One participant had a concern that when users were left on their own they could not operate the visualization. System Requirements One participant was concerned that a high end gaming PC is needed to operate the system.
Light Visualization Two participants stated that they did not need the internal light visualization in its current form. They would like to have the ability to relocate lights. Surroundings Two participants stated that the visualization is not connected to the actual surroundings making the displayed surroundings feel too abstract and unnatural. General Feedback The participants were giving the option to give general feedback. Season Visualization One participant suggested a feature to visualize seasons to even more emphasize how the building would look throuout the year.
Glass Material One participant stated that he did not like the relections of the glass material; he would have expected to see himself in the glass when walking near the glass window. This way he had no way to distinguish between an empty space and a window. The user suggested adding a targeting cross hair that indicates if the use function is available or not. Observations Here are some observations on how the participants were using the prototype.
The remaining 3 participants had never before played computer games and needed an introduction to the controls. Performing actions using the E-Key was familiar due to similar use in video games. On the other hand, opening the menu with the R-Key was unfamiliar to all participants and had to be explained. Users expected to be able to open the menu using the escape key. No participant utilized the arrow keys for navigation. Personal All participants reacted to the prototype in a rather personal way; they immediatly drew comparisions with their current living arrangements or work environment.
It was a nice gimmick without any practical use. However, when asked about their current living situation it was revealed that they lived in their home for more than 30 years and did not have any intension of relocating, Participants of age , were housewives, primarily wanted to use the prototype for interior planning. Discussion The study shows, even though the sample size is small, a clear tendency that the users were very engaged in the visualization. None of the participants had major dificulties using the system.
Minor usability features could be added to create a smoother user experience. The viewers were neither aware of the artiicial nature of the sky visualization nor of the lighting design aspects. Average users take lighting for granted. The users were not aware of the lighting system speciically designed for architectural light visualization. The biggest inluence on the perception of the visualization appeared to be dependent on the plan to relocate or invest in a property. If the user is happy with his current living arrangement and has no interest to invest, the viewer has no interest in the visualization.
The prototype gives the user the ability to control the visualization and view the building from any angle in different times of day. This tool would make the visualization a powerful marketing tool to appeal to the average user. Summary The project analyzed the capabilities of using Unreal game technology to be applied in the ield of architectural visualization. Maya was used to convert architectural data into UDK assets. This required the creation of a plugin to streamline the worklow. The aim was to lower the entry barrier for professionals and increase eficiency, as well as allowing rapid prototyping of 3D visualizations.
A modular UDK architectural visualization framework was created with following features: coniguration of the UDK to support architectural visualization, exchangeable environments, time of day visualization, interior lighting, basic architectural materials, and interface for architectural interaction. The modular nature of the framework allows components of the framework to be easily updated and extended. The core of the framework can be used as the foundation for different types of visualizations not limited to architectural visualization. Using the Maya plugin drastically reduces the time needed to create 3D assets; at the same time 3D artists require no explicit knowledge of the FBX worklow and the plugin is self- explanatory.
In combination with the framework this reduces the entry barrier to work with UDK.
Newcomers to the systems only require basic knowledge of 3D modeling, UDK, and optionally Flash, to create visualizations. The framework was used to represent the XY-house. The lighting features available in the UDK were further explored to be able to provide interior lighting.
A qualitative study was performed where users could interact with the visualization. The visualization was positively received. The participants were pleased to see a 3D visualized building instead of having to deal with technical drawings. They felt that this would be a good communication medium to conidently provide their ideas to the architect. Comments on the Lighting System We tried to focus on the specialized needs of architectural light simulation. One major aspect is the accurate representation of sunlight. For this there are two considerations: the light source is constantly moving, and the light changes its color.
This requires the interior bounce light to be updated dynamically. Full dynamic lights are needed to fulill these requirements. Lighting artists overcome this problem by introducing additional light sources that visualize bounce light . However, this technique only creates an approximation and is not based on a simulation model used for architectural light planning. Due to the restrictions of the freely available UDK, it cannot be extended with external code to support a full dynamic lighting model.
However, the next release of Unreal Engine 4 will support spherical harmonics  that are used for dynamic indirect light calculations. At this time the engines do not support full dynamic lights, and other features available in UDK are missing. However, they provide open access to the engine enabling programmers to implement missing features. Future Work UDK is a gaming platform that can be used for serious research .
Already now in the ield of military and medical research game technology is used for training purposes. The prototype framework we created covers the basic features needed for architectural visualization. More advanced features could be added to the system, like a weather system, or adding animated characters and vehicles to create a living city visualization. The time of day feature could be extended to a time of year visualization allowing the user to switch between various seasons of the year.
UDK has the possiblity to utilize different types of input devices. Katsarou , G. Zeilig , M. Plotnik , A. Gotlieb , S. Fountoukidou , J. Ham , D. Athanasiou , A. Mariakaki , D. Comandicci , E. Sabatini , W. Nistico , M. Plank , I. Gkatziaki , S. Mills , S. Diplaris , I. Tsampoulatidis , I. Boididou , S. Middleton , Z. Jin , S. Dang-Nguyen , G. Boato , I. Kompatsiaris , " Verifying information with multimedia content on Twitter ", Multimedia Tools and Applications : Akhgar , P. Bertrand , C. Chalanouli , T. Day , I. Fragkiadakis , H. Gibson , D. Kavallieros , E. Kermitsis , I. Kompatsiaris , E.
Kyriakou , G. Leventakis , E. Lissaris , S. Mille , D. Myttas , T. Tsikrika , S. Vrochidis , U. Tachos , K. Iliou , G. Kalpakis , T. DOI: Gialampoukidis , A. Moumtzidou , D. Liparas , T. Kompatsiaris , " Multimedia retrieval based on non-linear graph-based fusion and partial least squares regression ", , Multimedia Tools and Applications, Vol.
Ioannidou , E. Poularakis , K. Briasouli , I. Kompatsiaris , " Efficient motion estimation methods for fast recognition of activities of daily living ", Signal Processing: Image Communication, T. Stavropoulos , G. Konstantinidis , T. Stathaki , V. Argyriou , N. Zygouris , K. Ntovas , D. Giakoumis , K. Votis , S. Doumpoulakis , S. Segkouli , C. Karagiannidis , D. Tzovaras , M. Tsolaki , " A preliminary study on the feasibility of using a virtual reality cognitive training application for remote detection of mild cognitive impairment ", Journal of Alzheimer's Disease 56, no.
Chantas , S. Karavarsamis , S. Kapoutsis , S. Chatzichristofis , E. Barmpoutis , C. Zioga , A. Kamas , K. Patsiaoura , N. Sanchez-Rico , P. Garel , I. Notarangelo , M. Quintana , G. Hernandez , S. Asteriadis , M. Popa , N. Vretos , V. Solachidis , M. Burgos , A. Segkouli , I. Tzovaras , I. Lazarou , C. Karagiannidis , F. Vlachos , M. Cheimariotis , Y.
Chatzizisisc , V. Koutkias , K. Toutouzas , A. Giannopoulos , M. Riga , I. Chouvarda , A. Antoniadis , C. Doulaverakis , I. Tsamboulatidis , I. Kompatsiaris , G. Giannoglou , N. Tsolaki , E. Karathanasi , I. Dovas , E. Verykouki , A. Tsolaki , K. Adam , I. Kompatsiaris , Z.
Ota , M. Dao , V. Mezaris , F. Manakos , K. Chatzopoulos-Vouzoglanis , I. Gkinis , Z. Petrou , E. Stylianidis , G. Manakos , C. Karakizi , I. Gkinis , K. Alexiadis , A. Chatzitofis , N. Zioulis , O. Zoidi , G. Louizis , D. Tzovaras , C. Karagiannidis , M. Zikos , S. Tryferidis , D. Ioannidis , M. Vidaurre-Arbizu , C. Martin-Gomez , S. Zuazua-Ros , D. Arena , A. Tsolakis , S. Krinidis , C. Ziogou , D. Kiritsis , " Human resource optimisation through semantically enriched data ", International Journal of Production Research, Vol.
Schoeffmann , W. Bailer , W. Vrochidis , K. Uwe Barthel , L. Rossetto , " Interactive video search tools: a detailed analysis of the video browser showdown ", Multimedia Tools Appl. Sansone , K. Conci , G. Boato , V. Apostolidis , C. Collyda , V. Mezaris , " A web-based tool for fast instance-level labeling of videos and the creation of spatiotemporal media fragments ", Multimedia Tools and Applications, vol.
Lalas , S. Nousias , D. Kikidis , A. Arvanitis , C. Sougles , K. Moustakas , K. Verbanck , O. Usmani , D. Tzovaras , " Substance deposition assessment in obstructed pulmonary system through numerical characterization of airflow and inhaled particles attributes ", BMC Medical Informatics and Decision Making, 17 Suppl 3 : , Ioannidis , P. Krinidis , G. Stavropoulos , D. Manakos , E. Technitou , Z. Petrou , C. Karydas , V. Tomaselli , G.
Veronico , G. Mountrakis , " Multi-modal knowledge base generation from very high resolution satellite imagery for habitat mapping ", , European Journal of Remote Sensing, 49, , DOI: Spampinato , V. Mezaris , J. Mezaris , M. Cristani , " Special issue on fine-grained categorization in ecological multimedia ", Editorial Pattern Recognition Letters, Elsevier, vol. Tzelepis , D.
Galanopoulos , V. Patras , " Learning to detect video events from zero or very few video examples ", Image and Vision Computing Journal, Elsevier, vol. Schinas , S. Kompatsiaris , P. Mitkas , " MGraph: multimodal event summarization in social media using topic models and graph-based ranking ", International Journal of Multimedia Information Retrieval, 5 1 , , Papadopoulos , K.
Bontcheva , E. Jaho , M. Lupu , C. Tsolaki , " Age-Induced differences in brain neural activation during emotional stimuli ", Journal of Neuroscience, Elsevier, T. Meditskos , S. Andreadis , K. Avgerinakis , K. Papadaniil , V. Kosmidou , A. Tsolaki , M. Tsolaki , I. Kompatsiaris , L. Crispim-Junior , V. Buso , K. Avgerinakis , G. Meditskos , A.
Briassouli , J. Benois-Pineau , I. Kompatsiaris , F. Kikhia , T. Andreadis , N. Karvonen , I. Kompatsiaris , S. Pijl , C. Multidisciplinary Digital Publishing Institute, p. Barmpoutis , A. Kitsikidis , N. Drosou , I. Kalamaras , St. Tzovaras , " Accessibility-based re-ranking in multimedia search engines ", Multimedia Tools and Applications, Vol. Axenopoulos , D. Rafailidis , Ge. Papadopoulos , E. Houstis , P. Endres , G. Mountrakis , H. Jin , W. Zhuang , I. Manakos , J. Wiley Jr , C.
Lazarou , A. Karakostas , T. Stavropoulos , T. Tsompanidis , G. Asteriadis , P. Kaza , S. Thermos , K. Dimitropoulos , E. Dimaraki , P. Tzelepis , Z. Mezaris , B. Ionescu , I. Boato , N. Sebe , S. Chatzichristofis , L. Doitsidis , J. Pinto , J. Braga , E. Kosmatopoulos , " Real-time adaptive multi-robot exploration with application to underwater map construction ", Autonomous robots, 40 6 , Baldi , I. Michailidis , E. Kosmatopoulos , " Occupancy-based demand response and thermal comfort optimization in microgrids with renewable energy sources and energy storage ", Applied Energy, , Michailidis , C.
Kosmatopoulos , E. Nassie , " Automated control calibration exploiting exogenous environment energy: An Israeli commercial building case study ", Energy and Buildings, , Tsolakis , D. Meskos , A. Kompatsiaris , J. Mochament , A. Tsakiris , D. Ferro , M. Girolami , D. Salvi , C. Mayer , J. Gorman , A. Grguric , R. Ram , R. Sadat , K. Giannoutakis , C. Chatzizisis , K. Riga , A. Antoniadis , Y. Fujinom , D.
Mitsouras , V. Koutkias , G. Cheimariotis , C. Chouvarda , I. Nakamura , F. Rybicki , N. Maglaveras , D. Tousoulis , G. Salamanis , D. Kehagias , C. Filelis-Papadopoulos , D. Barmpoutis , T. Koletsa , I. Kostopoulos , N. Koutkias , N. Tzovaras , N. Adamopoulou , A. Tzovaras , " A context-aware method for building occupancy prediction ", Energy and Buildings : Bikos , M. Karypidou , E. Stalika , P. Baliakas , A. Xochelli , L. Sutton , G. Papadopoulos , A.
Agathangelidis , E. Papadopoulou , Z. Davis , P. Algara , G. Kanellis , A. Traverse-Glehen , M. Mollejo , A. Anagnostopoulos , M. Ponzoni , D. Gonzalez , S. Pospisilova , E. Matutes , M. Piris , T. Papadaki , P. Ghia , R. Rosenquist , D. Oscier , N. Darzentas , D. Belessi , A. Hadzidimitriou , K. Koureas , A. Hardisty , R. Vos , D. Agosti , C. Arvanitidis , P. Bogatencov , P. Buttigieg , Y. Horvath , G. Gkoutos , Q. Groom , T. Kliment , U. Manakos , A. Marcer , K. Marhold , D.
Morse , P. Mergen , L.
Penev , L. Pettersson , J. Svenning , A. Mobile Comput. Tzovaras , O. Oikonomou , G. Liaros , K. Georgiadis , E. Chatzilari , K. Adam , S. Palaiokrassas , A. Voulodimos , K. Konstanteli , N. Vretos , D. Osborne , E. Chatzi , P. Moustakas , A. Barmpoutis , N. Giakoumis , E. Votis , K. Kaklanis , P. Biswas , Y.
Mohamad , M. Gonzalez , M. Peissner , P. Langdon , D. Jung , " Towards standardisation of user models for simulation and adaptation purposes ", Universal Access in the Information Society, Vol. Rogotis , C. Palaskas , D. Likothanassis , " Recognizing suspicious activities in infrared imagery using appearance-based features and the theory of hidden conditional random fields for outdoor perimeter surveillance ", Journal of Electronic Imaging, Vol. Kordelas , D. Alexiadis , P. Daras , E. Kaklanis , G. Tzovaras , " Modeling people with motor disabilities to empower the automatic accessibility and ergonomic assessment of new products ", Applied Ergonomics, Vol.
Izquierdo , " Content-based guided image filtering, weighted semi-global optimization and efficient disparity refinement for fast and accurate disparity estimation ", IEEE Transactions on Multimedia, Vol: 18, Issue 2, pp. Kosmidou , L. Hadjileontiadis , I. Mantziou , S. Kompatsiaris , " Learning to detect concepts with Approximate Laplacian Eigenmaps in large-scale and online settings ", International Journal of Multimedia Information Retrieval 4 2 , pp. Tsolaki , L. Hadjileontiadis , M. Epitropou , T. Bassoukos , K. Karatzas , A.
Karppinen , L. Wanner , S. Kompatsiaris , " Activities of daily living recognition using optimal trajectories from motion boundaries ", Journal of Ambient Intelligence and Smart Environments 7. Dasiopoulou , I. Cristea , I. Kosmatopoulos , " Intelligent energy and thermal comfort management in grid-connected microgrids with heterogeneous occupancy schedule ", Applied Energy, , Chatzichristofis , G. Sirakoulis , L. Doitsidis , E. Springer, Cham. Semertzidis , J. Novak , M. Lazaridis , M. Melenhorst , I.
Micheel , D. Michalopoulos , M. Strintzis , P. Patras , " Learning to detect video events from zero or very few video examples ", Image and Vision Computing Journal, Elsevier, , accepted for publication. Mariolis , S. Malassiotis , " Modelling folded garments by fitting foldable templates ", Machine Vision and Applications, Vol. Roupioz , H. Kramer , M. Bogers , R. Jongman , R. Lucas , V. Kosmidou , Z. Petrou , I. Padoa-Schioppa , M.
Adamo , P. Baldi , A. Karagevrekis , I. Kosmatopoulos , " Joint energy demand and thermal comfort optimization in photovoltaic-equipped interconnected microgrids ", Energy Conversion and Management, , Boulgouris , K. Dimitropoulos , N. Kehagias , A. Wanner , M. Rospocher , S. Vrochidis , L. Johansson , N. Bouayad-Aghae , G. Casamayor , A. Karppinen , I. Millee , A. Moumtzidou , L. Serafini , " Ontology-centered environmental information delivery for personalized decision support ", Expert Systems With Applications, Volume 42, Issue 12, 15 July , Pages — Kaltsa , A. Mezaris , N. Pittaras , I.
Manakos , T. Stathaki , " Remote sensing for biodiversity monitoring: A review of methods for biodiversity indicator extraction and assessment of progress towards international targets ", , Biodiversity and Conservation, 24 10 , Tzovaras , T. Tsakiris , M. Tsolaki , C. Semertzidis , D. Rafailidis , M. Michailidis , S. Baldi , E. Kosmatopoulos , M. Pichler , J. ISBN: Tsikrika , A. Moustakas , M. Petrou , " Activity related authentication using prehension biometrics ", Els.
Pattern Recognition, vol. Stathaki , C. Johansson , V. Epitropou , K. Bassoukos , J. Kukkonen , I. Kompatsiaris , " Fusion of meteorological and air quality data extracted from the web for personalized environmental information services ", Environmental Modeling and Software Journal, , Volume 64, February , pp. Chatzopoulos-Vouzoglanis , Z. Petrou , L. Filchev , A. Spampinato , G. Farinella , B. Boom , V. Betke , R. Dasiopoulou , G. Stevenson , G. Lucas , P. Blonda , P.
Bunting , G. Jones , J. Inglada , M. Arias , V. Adamo , R. Charnock , C. Tarantino , C. Jongman , H. Kramer , D. Arvor , J. Honrado , P. Nagendra , P. Mairota , C. Marangi , R. Dimopoulos , J. Honrado , M. Niphadkar , C. Tomaselli , M. Panitsa , C. Tarantino , I. Manakos , P. Blonda , " Satellite Earth observation data to identify anthropogenic pressures in selected protected areas ", , International Journal of Applied Earth Observation and Geoinformation 37, Zygouris , I. Lazarou , I. Chatzileontiadis , K. Tzovaras , A. Karakostas , C. Karagkiozi , T.
Dimitriou , T. Tsiatsios , S. Dimitriadis , I. Tarnanas , D. Dranidis , P. Heut , J. Martin-Gomez , M. Vidaurre-Arbizu , M. Eguaras-Martinez , S. Tsolaki , D. Kazis , I. Kosmidou , M. Wanner , H. Bosch , N. Bouayad-Agha , G. Casamayor , T. Ertl , D. Hilbring , L. Johansson , K. Kompatsiaris , T. Koskentalo , S. Mille , J. Mossgraber , A. Moumtzidou , M. Myllynen , E. Pianta , M. Rospocher , L.