Data Management

• Automatic data import from AWS S3.
• Upload of data into datasets
• Simple and Fast creation of image embeddings
• Data Indexing in high dimensional vector space for fast similarity search

Collection Querying / Similarity Search

• Query Collections with images from the database or by uploading a file
• Fast query times for similarity search
• Finds duplicates and similar images in a few milliseconds

Data visualization

• Comprehensive exploration of image embeddings through an interactive visualization
• Find clusters of similar images
• Analyze the dataset interactively by navigating a 2D-Vector Space
• Query images and see where similar images are located in the 2D-Vector Space

What is possible?

• Exploration of public Collection "Open Images"
• Query and Explore the collection by either uploading your own query images
  or navigating through the interactive visualization

What is not possible?

• User Registration disabled
• Dataset and Collection Creation disabled
• Import of new data disabled

Frontend

• React
• Typescript
• deck.gl for visualization
• Auth0 for Authentication

Backend - Web Application / Rest Server

• NodeJS / ExpressJS
• MongoDB

Workflow Orchestration

• Apache Airflow
• Python

Vector Embeddings and Data Analysis

• Python
• Tensorflow, Image Classification Models, Embedding Models
• VGG16 Vector Embeddings
• Milvus Vector Database
• RabbitMQ - Message Queue
• T-SNE, UMAP

Cloud

• AWS S3
• AWS Batch
• AWS EC2