How to Build Video Analytics for Live Streams Without Big Tech Resources
summary
A team from Sibur describes building video analytics for live streams from over 2,000 cameras across multiple factories, tackling tasks like production process monitoring and safety-compliance detection (e.g., harnesses, helmets). They explain how to run large-scale, real-time video analysis with far fewer compute resources than a big tech company, focusing on practical architectures and engineering trade-offs for constrained infrastructure.
AI-written summary · about
chapters (5)
- 0:00Introduction
- 3:16System constraints and challenges
- 13:54Local performance optimizations
- 25:37Architecture changes and results
- 38:11Q&A
related talks
53mProcessing streaming video for continuous monitoring
28mVision, Big Data and AI
47mBest Practices for Real-time Intelligent Video Analytics
29mApplied Custom Vision
36mVideo analytics at scale: Deep Learning, Computer Vision, ML & Spark
46mComputer Vision to Secure Your Surroundings with AI/ML Solution Built Using Open Source Tools at the Edge
40mDemocratise public agency user access to AI/ML on VAS
15mVideo Analytics with Zero-streaming Cameras