Enhancing Shopping with Machine Learning and More

When I started developing Heima, it was out of a need to simplify the chaos of family life. Between planning dinners, managing grocery lists, driving the kids to activities, and trying to remember birthdays, life was filled with sticky notes on the fridge, disappearing lists, and a collection of apps that never quite did what I needed. I wanted an app that could bring all of these tasks together into one seamless experience.

Over time, this simple idea evolved into the Heima app that it is today. With the latest update, we’ve introduced a few features that I’m particularly proud of. One of the highlights is the machine learning-powered barcode scanner that can pull prices directly from price tags. Balancing accuracy and performance on mobile devices was a technical challenge, but the results have been worth it. Shopping has never been so efficient.

I’ve also improved the authentication system by implementing OAuth for a more secure and seamless login. In addition, we’ve added new features like built-in chat to stay in touch with friends and family, loyalty card management so you never miss out on benefits, and a dark theme for those who use the app late at night (a feature I personally really appreciate as someone who often works after the kids go to bed).

For our international users, we’ve expanded language support to include German and improvements to French and Icelandic translations, making Heima more accessible to families around the world. On the technical side, the home screen has been completely overhauled, performance improvements have made the app faster, and we’ve fixed a number of bugs to make it more stable.

This is just the beginning. There’s a long list of features I’m working on, and your feedback is key to shaping the app’s future. Developing and maintaining an app like Heima as a solo developer is no small feat, but it’s been incredibly rewarding. My goal has always been to create something that makes life at home easier for families like mine—and yours. And when I say WE, I mean my wife and me. She is my biggest supporter and tester, providing feedback on everything I create.

If you haven’t tried Heima yet, now is the perfect time. Let me know what you think and stay tuned for more exciting updates.

• Google Play
• App Store

Behind the Scenes: Building the Price Detection Feature

One of the most unique features of Heima is the ability to retrieve prices from price tags using the barcode scanner. This functionality, which may seem simple, required significant work behind the scenes. Here’s how it came to be:

To train the system, I collected around 500 images of Norwegian price tags from local stores. Collecting this dataset was an important first step, as it provided the real-world examples needed to teach the model to accurately identify price information.

I used Roboflow’s image segmentation tool to process the images and created a project called the Price Tags Computer Vision Project. This tool allowed me to label the images and segment out areas of interest, focusing on the prices and their context. You can see the project here: Price Tags Computer Vision Project.

Once the segmentation was complete, I trained a model using YOLOv11, a state-of-the-art object detection framework. This model was then exported in two formats to ensure cross-platform compatibility:

• Apple’s .mlmodel: Optimized for iOS devices.
• TensorFlow Lite: For Android users, providing a lightweight and efficient solution.

The integration of these models into Heima ensures that price detection is not only accurate, but also works well on mobile devices. This feature is groundbreaking for grocery retailers, allowing users to scan price tags directly from their shopping list and retrieve price data in real time.

For those interested in the technical details or who want to experiment with their own price detection projects, you can find more about the implementation and access the source files here: Shop Price Tag Project.