Video Analytics for mere mortals with NVIDIA DeepStream SDK

on YouTube ↗53m

" summary

Demonstrates building real-time video analytics pipelines with NVIDIA DeepStream SDK using graph composer and Docker images, running on GPUs (Tesla T4, RTX 2080/3080) or jetson devices. Shows concrete pipelines: RTSP/file source → batched preprocessing → detector → tracking IDs → onscreen boxes and message broker output. Integrations push JSON metadata to Azure IoT Hub or Kafka, enable on-device smart recording and model swap, and count objects crossing a line. Demos run end-to-end but docs are thin and some runtime quirks appear.

AI-written summary · about

" chapters (4)

1. 0:00Start
2. 1:51Trivia Time
3. 2:57Computer vision use cases
4. 12:59Development workflow

" related talks

• 47mBest Practices for Real-time Intelligent Video AnalyticsEkaterina Sirazitdinova · GOTO Copenhagen 2021
• 36mVideo analytics at scale: Deep Learning, Computer Vision, ML & SparkClaudiu Barbura · Global Artificial Intelligence Virtual Conference Sep 2020
• 45mExtracting real-time data at the edge with SageMaker & NVIDIA GPUsRyan Vanderwerf, Eddie Weill, Yuxin Yang · Amazon re:MARS 2022
• 17mAI/ML demo derbyZack Akil, Kaz Sato, Markku Lepisto · Google I/O 2021
• 57mDrones & AI - What's all the buzz about ?Adam Stephensen · NDC Oslo 2019
• 53mProcessing streaming video for continuous monitoringElizabeth González · WeCode Fest 2020
• 51mVideo Analytics for Football gamesSven Degroote · Devoxx Belgium 2019
• 53mVideo Object Detection at ScaleReza Zadeh · O'Reilly Artificial Intelligence Conference 2017 - New York, New York