High quality Virtual Reality for Architectural Exhibitions

Anette Kreutzberg11The Royal Danish Academy of Fine Arts, School of Architecture1anette.kreutzberg@kadk.dk

This paper will summarise the findings from creating and implementing a visuallyhigh quality Virtual Reality (VR) experiment as part of an internationalarchitecture exhibition. It was the aim to represent the architectural spatialqualities as well as the atmosphere created from combining natural and artificiallighting in a prominent not yet built project. The outcome is twofold: Findingsconcerning the integration of VR in an exhibition space and findings concerningthe experience of the virtual space itself. In the exhibition, an important aspectwas the unmanned exhibition space, requiring the VR experience to beself-explanatory. Observations of different visitor reactions to the unmanned VRexperience compared with visitor reactions at guided tours with personalinstructions are evaluated. Data on perception of realism, spatial quality andlight in the VR model were collected with qualitative and quantitative methods attwo different occasions and setups after the exhibition, both showing a highdegree of immersion and experience of reality.

Keywords: Virtual Reality, Oculus Rift, GearVR, Exhibition display

INTRODUCTION AND BACKGROUNDAsapart of theexhibition seriesWorldArchitecture atthe Danish Architecture Centre [10] (DAC) in Copen-hagen, the exhibition 'World Architecture - Snøhetta'[13] opened to the public in June 2015. The exhibi-tion, created by Snøhetta [12] and Danish Architec-ture Centre, ran until September 27th 2015 and had13.000 visitors. Experimental use of virtual realitywasimplemented in the exhibition to demonstrate stateof the art communicative methods and tools as wellas expand the range of media used to present archi-tecture in exhibitions.

The experimental VR model let the audience ex-perience the vast space of the Central Hall in KingAb-dulaziz Centre for World Culture [11] in a 1:1 scale asif they were actually present in the building complex

situated in Saudi Arabia, well before the buildingwasinaugurated.

Architectural representation and experi-encing architectural spaceIt is not enough to see architecture; you must expe-rience it (Rasmussen 1959). There is a discrepancy inthewaywe sense architecture and its representation.For an architect it is important to be able to preciselyexpress oneself andmake sure that visions and inten-tions are fully understood. Architectural represen-tations are primarily drawings or images, previouslyhandmade and nowadays mostly computer gener-ated from 3Dmodels, but it is still 2D representationspresented either on paper or on screen. Although of-ten supplemented by scale models to touch, the 2D

VIRTUAL REALITY - Volume 2 - eCAADe 34 | 547

representation remain the primary and purely visu-ally focused method of representing architecture.

Adding interaction or motion to screen imagescan help conveying space (Kreutzberg 2011) butthere is still a difference in envisioning the real spacewithin a frame of an image or animation and experi-encing the spacewith your body in the realworld. Ex-periencing architectural space is a multi-sensory ex-perience involvingmore thanmere vision (Pallasmaa2005).

VRhas thepotential of bridging thegapbetween2D visualization of space and the bodily experienceof space in the real world by placing the viewer in thevirtual world, making the viewer part of the virtualworld and therefore experiencing the virtual worldwith bodily senses.

Virtual spacePrevious studies have found VR displayed on Ocu-lus Rift DK2 [9] Head Mounted Display (HMD) tobe a promising media platform for visualizing anddemonstrating complex spatial 3D models, espe-cially for non-experts untrained in reading technicaldrawings (Kreutzberg, 2015).

VR is still an emerging medium, where hardwareand software are improved and refined constantly,many of which are still only available in developerversions. Strict limitations of the models concerningpoly count, texture size, lights and screen effects ap-ply, to run a stable and comfortable VR application.

Some issues are acknowledged and important toaddress especially concerning simulator sickness inVR (Moss, 2011). It is of major importance to preventfeelings of disorientation and nausea, and to protectusers from motor-visual functioning issues after use,especially at a public exhibition. Creating a stable ex-ecution with sufficient frame rate and limited drawcalls is the challenge.

The experience of realism in architectural VRmodels relates to the use of real world model size,detailed surface properties, shadows and reflectedlight. They are all important for experiencing and un-derstanding the architectural space but especially for

sensing the intended atmosphere, light is important.Real-time rendering of reflected light and soft shad-ows are very computational demanding and requireextensive optimization of the VR model for use withcontemporary available HMD's.

METHODSPhenomenological studies were made several timesduring the exhibition period, observing the be-haviour of the visitors in the area with the VR HMDat a distance. Differences in behaviour of groups, sin-gles and groups on guided tours were observed.

Survey at cultural eventAfter the exhibition period a cross-sectional anony-mous survey (Creswell 2014) was conducted at Kul-turnatten [2] collecting data [6] from visitors (N=72)at a cultural event displaying the KACWC VR model.Questions relevant for this paperwere focusedon theexperience of realism, image quality, degree of vi-sion and comfort as well as including a comparisonof viewing the VRmodel inOculus DK2 and as stereo-scopic 360 °rendered image in GearVR [8].

Qualitative empirical studyTo get a more thorough understanding of the ex-perience in the KACWC virtual space, data was col-lected on sensory, bodily and atmospheric experi-ences through video-recorded Walk-and-talk inter-views with nine participants, inspired by SensoryEthnography methods (Pink 2009).

The interviews were arranged after the exhibi-tion and the cultural event, and were conducted in ameeting roomat anarchitect's office. The interviewerfollowed along on a computer screen andmade con-versation about the sensuous experiencewhile it washappening. The interviews lasted between 8:20 min-utes and 13:35 minutes each. (Hermann 2016.)

After the interviews, the participants sum-marised and discussed their individual experiencesin two focus group discussions.

548 | eCAADe 34 - VIRTUAL REALITY - Volume 2

Exhibition design - practical setupThe exhibition Snøhetta: World Architecture was or-ganised in three rooms each representing the exhibi-tion subtitle: People Process Projects. The VR exper-iment was placed in the first room representing thepeople and their work environment. An Oculus RiftDK2 HMD was mounted on an acrylic box discretelyplaced on the wall, representing the very long officeworking table. The connected laptopwas hidden be-hind the wall and not visible to the audience. A smallinstruction label was placed in the box. See figure 1.

Figure 1The HMD.

TheHMDwasmodified to be used handheldwithouthaving straps tied to the head for two reasons. Firstly,a hand held experience restricts the movement ofthe upper body avoiding vigorously turning and sec-ondly, it places the hands "out of sight".

Some visitors had previous experience with theOculus DK2 or had seen it demonstrated online, andthey unfortunately wanted to use the heads traps,

and tried to tear the straps down that were securedon the side of the HMD.

Proper visual instruction as to how to hold theHMD with both hands instead should have beenmore visible. The comfort of not having the HMDstrapped to the face was obvious though, and peo-ple wearing glasses could actually use the HMDwiththeir own glasses.

VRmodelThe VR model was created by converting a Rhino [3]solid model to several mesh models in 3dsMax [4]and importing these mesh models into Unity [5] fortexturing, lighting and programming of interactionand finally output the complete model to an exe-cutable file.

Converting a huge conventional 3D Rhino archi-tectural solidmodel representing 100.000m² to a lowpoly mesh model suitable for VR is not an easy task.All unwanted details were deleted prior to the meshconversion and still the final optimised mesh modelhad more than 2 million triangular polygons. It wasnot possible to optimise all the geometry automati-cally without at the same time changing the appear-ance of the original architecture. Several parts of themodel were therefor completely remodelled to opti-mise and at the same time to lay out the topology forclean UV mapping of texture maps and light maps.All mesh optimizing and UV mapping were done in3dsMax. See figure 2.

Figure 2Central Hall in KingAbdulaziz Centrefor World Culture,Screen View.

VIRTUAL REALITY - Volume 2 - eCAADe 34 | 549

Lighting was created in Unity with PrecomputedReal-time Global Illumination with Forward Render-ing in combinationwith BakedGlobal Illumination. Itallows for the use of hardware techniques like 'multi-sample anti-aliasing' (MSAA) which can have a greatimpact on image quality.

The different light sources were adjusted in in-tensity, direction and colour in an attempt to bestrepresent the combination of daylight and artificiallight in the architecturalmodel.TheUnity Profilerwasused throughout the process as a tool to analyse andensure stability in performance of the final built.

RESULTS AND REFLECTIONUsing VR in an exhibition space without personal in-struction turned out to be a larger challenge thanfirst anticipated. First of all the VR headset was notspotted by quite a lot of visitors - it blended in withthe wall displays too well! Secondly if spotted - quitesome bewilderment arose as to what was displayedand what to do. Most visitors found out to lookaround but not all tried to turn their body 180 de-grees.

Initially it was intended to allow for movementand exploration in the VRmodel, but concerns aboutsimulator sickness and the practical setup in the un-manned exhibition turned the experience into an ex-ploration from a fixed viewpoint.

Halfway through the 3-month exhibition period,an update was made to the model creating a loopwith fade-to-black between two different positionsinstead of displaying just one permanent viewpoint.The loop was combined with plan views indicatingthe active point of view. See figures 3, 4 and 5.

Looping between two positions did not work asintuitive as expected - fade to black was set to 4 sec-onds for a comfortable feel - but it was often inter-preted as a fault - that the installationbrokewhen thescreen turned black - and visitors took of the headsetbefore they could realize it was a transition into thenext position.

Many visitors completely lost track of the orien-tation in real space in the exhibition, and were quitedisoriented when taking off the HMD again.

Figure 3KACWC plan view A.

Generally, women seemed more focused on explor-ing the architecture, sensing the light and general at-mosphere accepting the immersion of being some-where else while looking through the VR glasses,whereas many men moved their heads and bodiesvigorously to test the latency and precision of thetechnology behind the player.

There was a difference in the way people actedif they were in a group, with a lot of audio to align tothe realworld, and the single visitorwithout any com-forting companions. The latter tended to be gentlerin the movements and were examining the headsetafterwards - and reading the instructions on the wallcarefully before they tried the experience.

Most people tried to take a step or twowhile be-ing immersed in the VR environment - and it tooksome tries to realize that locomotion was not possi-ble.

On guided tours, the visitors were introduced tothe VR setup, and they could understand the loop-feature and were generally very impressed. Explana-tions as to what the model actually showed was alsoexplained by the guide.

It was a popular task in groups to film or photo-graph each other wearing the HMD.

Cultural eventAt the cultural event posters with renderings, draw-ings and text gave visitors an opportunity to learnmore about the architectural project displayed in VR.Visitors were guided in their VR experience and spentapproximately 5 minutes in total testing the differ-ent HMD's. The survey results (N=72) were very pos-

550 | eCAADe 34 - VIRTUAL REALITY - Volume 2

Figure 4KACWC position A.

Figure 5KACWC position B.

itive with 31% of survey respondents experiencingthe realism tobe "Fantastic" andanother 43%respon-dents experiencing the realism to be "Good". Degreeof vision and comfortwere experienced as "Fantastic"by 53% and 44% of survey respondents respectively.See figure 6.

Visitors used Oculus DK2 to view the looping VRmodel from the exhibition and had the opportunityto compare it with a pre-rendered stereoscopic 360°image of the same VR model (in this example ren-

dered from within Unity 3d game engine in a ratherblurry resolution!) displayed inGearVRon a SamsungGalaxy S6 [7]mobile. The comparisonwas interestingbecause the high quality VRmodel used static pointsof views, and did not utilised the possibility of move-ment. The same static point of view in apre-renderedstereoscopic 360° rendering can be rendered directlyfrom the render software without the need of timeconsuming optimising for real-time rendering in agame engine.

Figure 6Experience of theroomwith OculusRift DK2.

VIRTUAL REALITY - Volume 2 - eCAADe 34 | 551

Figure 7Experience of theroomwith GearVR.

Difference in experience of realism and image qualitywas noticed though including the smaller view frus-tum inGear VR (100°) compared toOculusDK2 (120°).See figure 7.

With the use of high quality renders, the GearVRdoes have great potential - it is cheap, portable, andyou can store many 360-renderings on the mobile.

Walk-and-talkThewalk-and-talk interviewswere conducted duringtwo days in November 2015.

Sixteen questions in three categories were in-cluded in the interview guide, asking for commentsranging from purely descriptive to emotional and in-tuitive reflections.

Participants were architects (4) and users of ar-chitecture (5) and the two groups had very differentapproaches to VR. The architects reflected mostly onthe representationof architectural spacewhereas thearchitect-users reflected on the virtual space and notthe representation of space.

Mass and volume.All participants described the dif-ferent components of the space, especially the con-trast between organic volumes emerging from regu-lar straight ceilings and walls. Most participants alsocommented on the large number of columns scat-tered in the hall.

Materials. The participants perceived most of thematerials correct. They identified wood, metal, pol-

ished stone, glass and concrete, only the rammedearth (walls) was generally confused with other ma-terials like cork, wood or brick. One exception was aparticipant confusing the polished stone floor with acarpet.

Light.Many participants described in detail differentlight settings, contrasts and emphasized the naturalor artificial appearance.

One architect commented on the different lightsetups in the two positions in the VR model:

(Pos. B) "I can see they have been working withartificial lighting in the ceiling. I am not particularlymad about it, if I may say so. I think that the skylightat the end of the room is impressive though, but itfeels like some daylight is missing, compared to theother place I was before.

(Pos. A) Here there are the same ceiling lightsas in the other place, but here it is much lighter ina pleasant way. It is because I can see that the lightcomes from above. There is a place by the stairswhere some light is, it could be daylight."

In thefinal focusgroupdiscussionher conclusionwas:"I think what worked best in VR was the repre-sentation of the light. [...] You got a great feel forthe character of light in VR, and you could experiencethe various light settings, daylight, reflected light andvarious types of artificial lighting. It is difficult to rep-resent light in an image. Here I really think it [VR] haspotential." (AN, Interview C)

552 | eCAADe 34 - VIRTUAL REALITY - Volume 2

Figure 8Videorecording ofhaptic impulsescausing unintendedmoving in realspace.

Haptic impulses. All participants expressed an urgeto explore the virtual space, especially "to look pastthe edge and down". The visual experience madethem forget the fact, that it was a stationary repre-sentation, and they tried to walk towards the pointsof interest. See figure 8.

Implicit sensory perception. "I have a feeling that ifI touched the pillar in front of me, I would feel stone.Cold against thehand. I clearly recall the surface fromplaces I have been before." (SM, Interview B3)

The bodily dimension. "I'm surprised at how wildthis is. I can look up, and I can turn around. Whichis a full 360 degrees! I can look down." (JD, InterviewB5)

Eyeheight. The eye height in the VRmodel was fixedand set to 165cm. Individual calibration of eye heightis possible with Oculus DK2, but since it was the ex-hibition VR model that was being tested, the fixedeyeheightwaspreserved. All participantswere in theheight range of 165cm.-180cm.

"The simple and tangible, such as standing up,means a lot. That you experience the space standing,instead of sitting at a conference table and lookingdown at a drawing or looking at a screen. [...] Youachieve a completely different physicality by stand-ing up and turning your body. It gives a physical feel-ing, which you cannot obtain in other ways. Otherthan by being in a room in the real world." (JJ, Inter-view C)

Atmosphere. For fully calling forth an atmosphere inVR, light andmateriality seemnot to be sufficient; themodel must contain furniture, people and other arti-facts to have a real feel of life to it.

CONCLUSIONReconstructing and optimising the very large Rhino3D model for use in VR was very time-consuming,but resulted in a stable execution including real-time rendered light. The model was on display inthe Snøhetta: World Architecture exhibition for 3months.

The integration of VR in the exhibition was too

VIRTUAL REALITY - Volume 2 - eCAADe 34 | 553

discreet in the unmanned area; many visitors didnot discover the HMD at all. For those that did,they seemed to have an exiting experience, althoughmany did not understand what was on display - nei-ther the content nor the technology.

The automatic loop between two fixed views in-troduced halfway through the exhibition period wasnot as intuitive as expected. Sufficient instructionson what to expect and why, especially concerningduration and (limited) interaction, proved to be veryimportant for a positive VR experience.

The perception of spatial and lighting qualitiesin the VRmodel worked very well and was examinedthoroughly atwalk-and-talk interviews after the exhi-bition period. The comment from a user: "It is almosttoo real," explains the urge to walk around and ex-plore the VR environment although interaction wasrestricted to fixed viewpoints. Since virtual reality is amedium that attempts to replicate one's experiencein the physical world, users are likely to have an ex-pectation that they will be able to interact with thatvirtual world in the same ways they do outside of it.

User responds indicate that for perception of at-mosphere in VR, light and materiality are not suffi-cient alone; the model must also contain a certainamount of furniture, people and other artifacts.

Further studies should extend the experimentswith representation of natural and artificial light inVR and should include experiments with 360° stereo-scopic renderings for faster iterative work.

ACKNOWLEDGEMENTThe Danish Arts Foundation [1] funded the develop-ment of the VR model.

REFERENCESCreswell, JW 2014, Research design: Qualitative, quanti-

tative, andmixedmethods approaches, Sage Publica-tions

Cutting, J 1997, 'How the eye measures reality and vir-tual reality', Behavior Research Methods, Instruments,& Computers, 29(1), pp. 27-36

Hermann, D 2016, Kroppen og sanserne i Virtual reality,Master's Thesis, University of Copenhagen

Kreutzberg, A 2011 'Game Engines as Dynamic Toolsin the Design Phase', First International Congress ofRetevitruvio 2011, Bari, Italy, pp. 1585-1593

Kreutzberg, A 2015 'Conveying Architectural Form andSpace with Virtual Reality', Real Time - Proceedingsof the 33rd eCAADe Conference 2015, Vienna, Austria,pp. 117-124

Moss, JD and Muth, ER 2011, 'Characteristics of Head-Mounted Displays and Their Effects on SimulatorSickness', Human Factors: The Journal of the HumanFactors and Ergonomics Society, 53(3), pp. 308-319

Pallasmaa, J 2005, 2008 edn., The Eyes of the Skin, Wiley-Academy

Pink, S 2009, Doing sensory ethnography, Sage Publica-tions

Rasmussen, SE 1959, 1980 edn., Experiencing Architec-ture, The MIT Press, Cambridge

[1] http://www.kunst.dk/english/[2] https://kadk.dk/kalender/kulturnatten-2015[3] https://www.rhino3d.com/[4] http://www.autodesk.dk/products/3ds-max/overvie

w[5] http://unity3d.com/[6] http://www.surveyxact.dk/[7] http://www.samsung.com/dk/galaxys6edge-and-gala

xys6/[8] https://www.oculus.com/en-us/gear-vr/[9] https://www.oculus.com/en-us/rift/[10] http://www.dac.dk/da/dac-life/udstillinger/2015

/snoehetta---world-architecture/[11] http://en.kingabdulazizcenter.com/[12] http://snohetta.com/[13] https://kadk.dk/en/news/ind-i-snohetta-forsknin

g-og-vr-briller-abner-op-arkitektur-det-er-bygget

554 | eCAADe 34 - VIRTUAL REALITY - Volume 2