Real Time Video Indexing Storage System

Purdue Research Team CAM2 Summer 2019 Project

CAM2 Image database team have been working on real time video indexing system. The project is currently work-in-progress with research paper aimed to be completed.

Source code: https://github.com/PurdueCAM2Project/CAM2ImageDatabase

Overview

The project aims to process incoming video streams from multiple IP cameras around the world, extract features from the frames, store the images as objects along with corresponding features, and make them ready for user query.

Key Technology

Database

• Vitess
  • A database clustering system for horizontal scaling of MySQL
• MinIO
  • MinIO is a cloud storage server released under Apache License v2, compatible with Amazon S3. As an object store, MinIO can store unstructured data such as photos, videos, log files, backups and container / VM images. The maximum size of an object is 5TB.

Object Detection

• YOLOv3

Design

Research Paper

Our research paper has been accpeted to COMPSAC 2020 conference.

• A Real-Time Feature Indexing System on Live Video Streams