Built by — Bakken & Bæck
A technology-driven design studio based in Oslo, Bonn, Amsterdam, and London. We define, design, and develop all things digital, in no particular order. With solid expertise in a wide range of design disciplines, we believe we can bring any idea to life — no matter the complexity. One of Fast Company’s Most Innovative Companies 2020.
Room Shuffle is a technical prototype for an app which aims to unleash previously untapped creative potential when it comes to your home. It’s a special assistant designed not only to help you furnish your space, but also to encourage you to open your eyes to unexplored possibilities.
Room Shuffle is trained on a collection of expert-created floor plans from which it learns about different room types and common patterns of furniture. For example a sofa tends to sit beside a coffee table, or a dining room comes with chairs. It learns rules for space around furniture, and how it tends to be positioned in a room. As the user, you would be able to use the app on completely empty rooms, or give it input on existing furniture pieces, doors, windows and lights. You choose the furniture to be placed, or leave it all up to the algorithm. Generated furnishings can be viewed live in three-dimensions (3D) directly in your space on your phone or tablet – allowing you to instantly witness how the suggestions would work according to your existing space.
'We hope that by both giving people personalised suggestions, and letting them see these live in their home, we will enable them to explore their own sense of space and style more, and feel more confident in making home furnishing choices.'
By feeding Room Shuffle with predefined and stylistically coherent room-sets of furniture, the algorithm trains a series of probabilistic models for the occurrence, placement, rotation and padding of furniture in rooms of different types. For a machine learning project of any kind, the behaviour of the final system depends a lot on your training data. Bakken & Bæck knew that if the app was trained exclusively on a three bedroom family home, the system would not perform as well on, say, a studio apartment, a cottage, or a loft conversion.