Designed To Fit

Designed to FitDesigned to FitChallenges of Interaction Design for Challenges of Interaction Design for Clothes Fitting Room TechnologiesClothes Fitting Room Technologies

Bo Begole, parc

Takashi Matsumoto, Keio University

Wei Zhang, Oregon State University

Nicholas Yee, parc

Juan Liu, parc

Maurice Chu, parc

HCI International, July 2009

Implicit InteractionImplicit Interaction Conventional systems: User

initiates interaction and drives the system

Implicit Interaction: System perceives state of user and situation and acts on users’ inferred goals

– Simplifies user experience– Little or no training required

ResponsiveMirror

Responsive Technologies InformationRecommendation

ClothingRecognition

Psychographic Profiling

[ICDSC 2008][IUI 2008, HCII 2009][CHI 2008]Magitti

Related PARC Research:also:Human-Robot Interaction

Multi-party conversations

Camera-based tracking

MIT Media Lab

Xbox Natal

Clothes ShoppingClothes Shoppingas an Information Seeking Activityas an Information Seeking Activity

Browse

Filter

Evaluate

Decide,Buy

Decide,Buy

In-StoreIn-Store OnlineOnline

Availability, cost, size, colors, texture, feel, fit, style trends, etc.

Related WorkRelated WorkApparel Fitting TechnologiesApparel Fitting Technologies

Virtual fitting technologies– Project image of clothing on image

of shopper or 3D model

– Helps pre-filter but ultimately clothing must be tried on

Fitting room technologies– Detect clothing items and retrieve

information – price, colors, in stock– Record videos of fitting– Send to friends for comments– Project virtual clothing on mirror

– Don’t provide access to prior trials– Don’t show social context of fashion

Responsive Mirror Responsive Mirror Implicitly controlled vision-based system providing Implicitly controlled vision-based system providing information for “self” comparison and “social” contextinformation for “self” comparison and “social” context

The Future of … Dressing Rooms, BNet, Sumi Das

video (min1:25-2:02)

Privacy ConsiderationsPrivacy Considerations

What are the implications of introducing a camera to this semi-private setting?

Altman’s three boundaries– Disclosure – clothes fittings are typically only shared with

co-present shoppers, friends, family– Identity – Apparel trials are a time when a person is

experimenting with their “presentation of self”– Temporal – fitting sessions are usually ephemeral, not

preserved for future scrutiny

Social aspects– What image do I project when wearing these clothes?– What do other people who wear these clothes look like?

Formative Formative StudyStudy

Quick examination of privacy concerns

Participants– 12 males, 28-52 years

» Limited population» Clothes matching

algorithm only worked on men’s shirts

Within subjects analysis Three Conditions

– Mirror alone– Mirror with Previous Outfit– Mirror with Other people in similar outfits

Task– Select one out of 6 shirts for each condition (18 shirts total)

Results - OverallResults - Overall

Buying decisions were not different across conditions– The characteristics of the clothing mattered more than the

technology

Which condition did you prefer? (1=highest, 5=lowest)

– Previous Outfit plus Other People (M = 1.92) (not experienced)

– Previous Outfit (M = 2.00)

– Other People (M = 2.83)

– Plain Mirror (M = 3.25) (2 = 9.10, p = .03)

Which condition was more helpful?– Previous Outfit (M = 3.00)

– Other People (M = 2.5)

Results - DisclosureResults - Disclosure How bothered by people in the following groups seeing

images from the fittings?– (5=bothers me a great deal, 1=doesn’t bother me at all)

– Family (M=1.08) – Friends (M=1.50)– Stranger (M=2.08)– Co-worker (M=2.25)

– No significant effect regarding gender– Level of concern was significantly higher for bad shirts (M=3.0)

versus good shirts (M=1.42) (p = .001)

Implication: Access control at just two levels:– Friends and Family– Co-workers and Strangers

not sig. diff’t

not sig. diff’t

sig. diff’t

Results - IdentityResults - Identity

How often do you think about– (5=Always, Often, Sometimes, Seldom, Never=1)– Someone you know who might like these clothes (M=2.67,

SD=0.98)– How similar to what other people you know are wearing

(M=2.92, SD=0.9) – How similar to what other people you don’t know are

wearing (M=2.33, SD=0.98)

Implication: Providing information on what other people are wearing would be useful sometimes

Results – TemporalityResults – Temporality

Would you want to remove images in the future if your tastes change? (5=Definitely, 1=Definitely Not)– (M=3.08, SD=1.16) (closest to Possibly)

Preferred periods of time– within 3 months (5 participants)– within 1 year (5 participants)– never (2 participants)

Implication: systems can remind users at 3 month and 1 year period to review their image record

Summary &Summary &ConclusionsConclusions

Physical apparel shopping requires different information than online shopping

– Apparel fit, feel, drape, texture, translucency, etc.

Privacy is a concern but not a block – understanding user concerns in Altman’s three

boundaries can help system designers

Future Directions– Countertop version for eyeglasses, jewelry,

hats, makeup, hair, …– Front-camera tracking algorithms– Adding sales support to the system

Contacts welcome!– [email protected]

Counter-top Responsive Mirror

The end