PROJECT TYPE
Mobile
Web
Enterprise
AI-based Interior Design Solution
Sometimes the designs in our head don’t always look good in real life. There’s where interior design apps come in for all the experiments.

Our AI-based interior design application lets you try out wallpapers in your home before buying. Just need to turn on a phone camera, select a suitable type and material and a new interior design will pop up in a photo of your space! With a few clicks neural network technology helps a retail salesperson and customers to make a choice. It is a clear and really good business model here – selling people the real stuff, once they’ve seen it virtually in their room.


TASK

Development of an AI-driven iOS application, web demo and widget for instant wallpapers modeling.


TECHNOLOGIES
swift
python
Solution

Mobile iOS application created use neural network to show the customers new interior design ideas in just a few clicks.

It works by the following algorithm:

After receiving a photo the application recognizes the ceiling, floor, furnishings and decorations. Then, the user has a chance to apply his favourite wallcovering options on the photo and the existing wallcovering is replaced by a new choice. You can try different types and colours of wall until you satisfied with the results and finished the room repair.

Augmented Reality mode can be used for more accurate overlay. 

For example, when a user wants to take a series of pictures for a room, for ex. 20 pics. It's easier to use the augmented reality mode and place the room manually, and then take pictures. Because automatic marking does not handle any photo - 20 photos will be processed in the first for a long time, secondly, not very accurately.

How is the room placed manually?

The user turns on the camera, he comes up one by one to each of the walls of the room. The wall is automatically recognized and a marker is added to the wall. After all the walls are marked you can take photos. If the wall is not recognized automatically, the user will need to place the marker manually.

To make the application run according to the algorithm, the following features were implemented:

  • Visualization tool for previewing how rooms look in different colors;
  • Built-in measurement feature;
  • Preserving the shadows, scale and perspectives of the room;
  • Distinguishing elements like pillars, wall niches, arches etc.
  • A set of wallcovering options available including colours, textures and geometric patterns;
  • Eventually, the ability to save the final design for future reference and share it with friends.

Integration options:

  • An internet store plug-in;
  • A workspace or self-service terminal in the offline store.
Web demo for stands works just like a demo with pre-prepared results. Web widget works just like a mobile application but without augmented reality.

Technologies used:
Swift
Python
Keras
numpy
cv2
Docker

Explore more projects.