PhotoMind

Stop saying "I'll look at my pics one day." Use AI-powered image classification to get there faster.

Comment

Inspiration

I've been there - frustrated at being out of storage. I had thousands of images, screenshots, and videos on my phone. Yet it's hard to let go of any, as I tell myself "one day I'll look back." But I already know this isn't true. Going through an unstructured mess isn't worth the nostalgia.

What it does

We built PhotoMind - the AI powered image classification service, designed to work on your photo gallery, which'll help you get your fix of nostalgia faster!

How we built it

We made use of Amazon Web Services extensively. Images are uploaded and stored securely in an S3 bucket. They are then tagged using AWS Rekognition, which assigns keyword tags to an image. These can be used to search for tags on an image. However, what makes us a league ahead of other photo apps is the ability to search with natural language, powered by Claude's AI.

Challenges we ran into

Ensuring that deep search was able to produce consistent results, as the limitation is the non-determinism of AI.

Accomplishments that we're proud of

We're proud of locating the exact service for assigning tags to an image, along with its confidence level. This helped us massively in building the rest of the app.

What we learned

DynamoDB and its streams are very powerful for allowing this app to have near instant access times and keep this incredibly scalable

What's next for PhotoMind

  • Bulk Upload, Supporting Google Photos, iCloud, etc as direct uploading
  • Ability to edit or delete tags to help search accuracy.
Share this project:

Updates