Memory Lane

Inspiration

We all have thousands of photos sitting forgotten on our phones, and we wanted to preserve those memories to look back on. We believed AI could help reconnect people with their memories. Pictures are one of the view things you can take with you through time, so having a digital and secure album for every memory keeps them special.

What it does

Memory Lane is an AI- =powered web app that automatically organises users’ photos into meaningful collections. When a user uploads an image, the system uses computer vision and natural language tagging to categorise them into personalised albums e.g. beach days, or family holidays. Each photo is automatically assigned smart tags, allowing users to instantly search for specific moments.

How we built it

We initially divided our responsibilities into two teams, front-end and back-end, with two of us working on the back and one on the front. The front-end was primarily built with React, Tailwind CSS and JavaScript. The backend is mainly python, using libraries such as FastAPI, SQLite, SQLAlchemy, Microsoft Azure's Computer Vision API and OpenAI. We hosted our site using Vite during development, and deployed it online using Railway.

Challenges we ran into.

The biggest challenge we had was implementing the backend with the frontend, due to developing them on different machines, the libraries and dependencies weren’t aligned, and our requirements files weren’t up to scratch. Also, referencing the database from the frontend, especially when logging in and uploading media was a struggle. We also struggled with things such as merge conflicts and branching. However, a few hours into the project we got the hang of it and ensured our code base was and remained organised, using branches for each feature, merging regularly and using reusable features.

Accomplishments that we're proud of

Developing a full-stack application as a team was a great experience to work on a larger development process than we were all used to. We quickly learnt how important it was to take a step back and look at the bigger picture and how we could all work together synchronously. We are all truly proud of our project and of each other, as we all did a much better job than we initially imagined we could. Implementing the APIs was a huge milestone since this was a new concept to us. DurHack has inspired us to do the hackathon in Manchester next week, and we will definitely continue developing this project and do more projects as a team.

What we learned

We learned how to integrate AI and computer-vision tools into a full stack platform for the first time. FastAPI and Azure’s Vision API, Railway and several libraries were completely new to us, so this project was a huge learning curve. We also improved massively in collaborative development, managing Git branches, resolving merge conflicts, debugging cross platform issues, and aligning the frontend and backend.

What's next for Memory Lane

We didn't have time to complete the facial recognition feature to categorise individuals. We planned to have a people page, where the user could assign a “person” a name. Which would then help optimise our searching, organising and album creating algorithms. We would like to add more tags and metadata such as geolocation, timestamps and context descriptions and add upload support for video and audio files. We’d also love to generate AI “Memory books”, a downloadable printable photo album.