The problem Indecore - Where technology meets design solves

The initial task on homeowners’ lists when renovating or redecorating their homes is to paint the walls. This simple project provides an easy way to give a room a fresh look and can even set the tone for the entire house. However, choosing the right paint colour can be overwhelming due to the endless options available, and a poor decision can be costly

To alleviate this problem, we come up with a robust system "Indecor" that enables homeowners to visualize how their homes would appear with a new coat of paint.This system is intelligent enough to ignore the complex objects within the house and accurately paint only the necessary wall sections, while still maintaining a realistic representation of the image.

The application uses computer vision and deep learning techniques such as the segment-anything to detect walls automatically and provide recommendations for wall painting and texture designs to the users. Additionally, the system provides a tool for users to select a wall and choose a color, and the AI model generates an image based on the user input.

Challenges we ran into

Finding the Right Image Segmentation Model
One of the biggest challenges we faced was finding the right image segmentation model that could accurately detect and segment the walls in a given image. We tested several models before finally settling on the "Segment Anything" model, which proved to be the most effective for our purposes.

Displaying Segmented Walls on the Image
Once we had successfully segmented the walls in a given image, we then faced the challenge of displaying the segmented walls on top of the original image. We experimented with several different techniques before settling on a combination of PIl and overlaying the segmented mask on top of the original image.

Showing Multiple Masks on a Single Image
Another challenge we encountered was the need to display multiple masks on a single image in order to show the different wall sections. We solved this problem by creating a custom overlay function that allowed us to display multiple masks on a single image in an intuitive and easy-to-understand way.

Recommending Wall Painting and Texture Combinations
Finally, we faced the challenge of developing a recommendation system that could suggest different wall painting and texture combinations to our users based on the segmented walls. We solved this problem by developing a machine learning algorithm that analyzed the colors and textures in the image and made recommendations based on established design principles.