Handy
A touchless control app recognising ten distinct hand gestures on-device, mapping them to phone actions without a visible camera preview.
View project overview , HandyCase study · Augmented reality
AR Furniture
A shopping app where the decisive feature is scale: furniture rendered at its true dimensions in the room it is destined for.
Published · In development
- Engagement
- Internal innovation project
- Platforms
- Android application
- Role
- Concept, engineering and 3D pipeline
- Stage
- Published, in development Verify before publishing
- Industry
- Retail
- Users
- Furniture shoppers
- Technique
- On-device augmented reality
- Platforms
- Android
- Role
- Innovation build
- Status
- Published
01The business situation
The question is never what it looks like. It is whether it fits.
Buying furniture online means guessing. A photograph tells you the colour and the style, and nothing at all about whether the piece will clear the doorway or swallow the room. Our innovation team took that as the whole problem: not a prettier catalogue, but a way to see a piece at its real size, in the actual space, before committing.
Core challenge
Render 3D furniture at true physical scale in a live camera view, without the rendering cost making the app unusable.
Before
Shopping from photographs and a tape measure
- Flat Photography
- Guessed Dimensions
- Manual Measuring
- Returns
After
The piece placed at true size in the real room
- Shopper
- Real Room
- 3D Model
- Decision
AR Furniture App
02The system at a glance
One app. Four connected parts.
- Product Catalogue Browse the available pieces
- AR Placement View Put the piece in your room
- Scale & Rendering True dimensions, live
- Manipulation Move, rotate and resize
Core PlatformOn-device AR with dimension-accurate models
- 3D Model Pipeline
- Surface Detection
- True-scale Rendering
- Object Manipulation
- Catalogue Browsing
- Performance Tuning
03Workflow 01
Point the camera. Put it down. Walk around it.
The app finds a surface, places the selected piece on it at its real dimensions, and lets the shopper move around it. Everything else in the product exists to make that moment convincing.
assets/portfolio/ar-furniture/workflow-place.png · 16 / 11
- Placed on real surfacesThe app detects the floor or table and anchors the piece to it rather than floating it in space.
- True to its dimensionsModels are rendered at the physical size of the real product, which is the entire point.
- Moved and resizedThe shopper can reposition the piece to compare placements before deciding.
04Workflow 02
Choosing the framework was the project.
The market offered several ways to render 3D on a phone and most of them either looked wrong or cost too much performance. The team built proofs of concept and chose on evidence rather than popularity.
- Frameworks compared with proofsSeveral rendering approaches were prototyped before one was chosen.
- Accuracy and performance togetherThe chosen approach had to render correctly without degrading the app.
- One model, then manyA single piece was taken end to end first, and the catalogue widened once the pipeline held.
assets/portfolio/ar-furniture/model-pipeline.png · 16 / 10
assets/portfolio/ar-furniture/catalogue.png · 1 / 2.04
05Engineering behind the product
The system behind a convincing placement.
Augmented reality is unforgiving: if the scale is slightly wrong, everyone notices, and the product loses its only reason to exist.
-
AR framework evaluation
Multiple rendering frameworks were prototyped before ARCore with Sceneform UX proved best on accuracy and performance together.
-
True-dimension rendering
Models are prepared and rendered at the real product’s measurements rather than an arbitrary scale.
-
Surface detection and anchoring
Pieces are anchored to detected floors and tables so they behave like objects in the room.
-
Object manipulation
Placed pieces can be moved and resized while remaining correctly grounded.
-
Native Android implementation
Built in Kotlin against the platform’s own AR capabilities for performance.
-
3D asset pipeline
Models are converted into the required format at correct dimensions before being added to the catalogue.
-
Performance tuning
Rendering was optimised so model quality did not come at the cost of a usable app.
06Technical architecture
Under the hood
Several rendering frameworks were prototyped before ARCore with Sceneform UX proved best for accuracy without a performance cost.
Android
- Kotlin
Augmented reality
- ARCore
- Sceneform UX
Assets
- 3D models at true product dimensions
07Delivery journey
How the app was delivered.
-
Concept
Innovation team frames the idea
-
Framework research
Proofs of concept across options
-
First model
One piece taken end to end
-
Business layer
Catalogue and app structure
-
Optimisation
Scale accuracy and performance
-
Catalogue expansion
More models, refined interface
-
Play Store release
Published and iterating
The app is published on the Play Store and development continues toward the next stage. Dates and durations are withheld until confirmed. Verify before publishing
08Where it stands today
What the app brings together
- Furniture rendered at true physical dimensions
- Pieces anchored to real detected surfaces
- Free movement and resizing in the live view
- A catalogue browsable before placement
- A validated framework choice backed by prototypes
- A published Android application
Related systems
Polar Attack
A tilt-controlled reflex game built on an open-source Java engine, with three difficulty levels and a published Play Store release.
View project overview , Polar AttackSome ideas have to be built before they can be judged.
Our innovation projects exist to prove a technique works, and we build on them with partners who want it.