Moment is a photo discovery platform driven by AI built for HackSC 2019

Winner of Best Use of Google Cloud Runner up for Euntrepeneurship Prize

One of the main issues with current photo albums is that it is difficult to search through masses of images efficiently. Moment attempts to solve that problem by providing two powerful search options: text-based search, and similarity search in order to help users locate specific photos, as well as navigate through their library through a novel experience.

Description

Moment lets you navigate through your photo library through similar photos and search through your photo library with text and images.

How it Works

  1. Users may upload images by dragging them into the application
  2. Images are passed through two separate deep learning models which provide tag data as well as feature vectors
  3. Images are indexed into Elasticsearch which allows filter by tag and vector similarity search using cosine distance
  4. You can explore your photo library through visually similar images, search and browse through intelligent albums using tags generated by the model

Technologies

Moment is powered by the following technologies:

  • Google Cloud (AI Platform, Vision API, Google Cloud Storage, Kubernetes)
  • XCeption Convolutional Neural Network
  • Elastic Search
  • Flask (Backend Server)
  • Isotope.js (Tiling for frontend)
  • Dropzone.js (For easy file upload)

Video Demo

Heres a video showcasing some of Moment's most powerful features.