WeAreDevelopers Logo WeAreDevelopers
Sign in
For You LIVE
World Congress 2026 Europe - Virtual Stage • Jul 2, 2026 • Session details

Challenges and Solutions for Efficient, Large-Scale Video Analysis

Magne Johansen

Magne Johnsen reveals that the fastest ML model still loses if the system bottlenecks. Discover how multithreading slashed video processing time from one hour to 40 seconds.

Pause
Mute Enter Fullscreen
#1 about 3 min

The role of pipeline architecture in video analytics

How the system surrounding a machine learning model dictates overall throughput and performance.

#2 about 2 min

Video analysis challenges for continuous animal monitoring

Scaling computer vision requires moving from short clips to continuous monitoring of complex interactions.

#3 about 3 min

Breaking down the three stages of video processing

Decoupling the workload into independent detection, tracking, and identification stages allows for targeted optimization.

#4 about 4 min

The mathematical barrier of sequential video processing

Why sequential processing and heavy local decoding make scaling to thousands of video hours impossible.

#5 about 4 min

Diagnosing hardware wait states through resource metrics

Interpreting CPU, GPU, and memory utilization together reveals whether accelerators are starved by pipeline inefficiencies.

#6 about 2 min

Custom Python profiling for pinpointing application bottlenecks

Building a custom annotation tool helps separate application logic overhead from external package execution time.

#7 about 3 min

Achieving quick baseline wins with cloud streaming

Streaming video directly from cloud storage and removing debugging code significantly improves baseline processing speed.

#8 about 4 min

Overlapping input operations and compute with multithreading

Decoupling video fetching and decoding from inference execution prevents the GPU from sitting idle.

#9 about 5 min

Implementing a producer-consumer buffer for video frames

Coordinating data flow with iterators and buffer queues smooths the handoff between background threads and the main loop.

#10 about 4 min

Scaling GPU utilization safely using Python multiprocessing

Distributing full videos across multiple CPU workers maximizes shared GPU capacity without introducing race conditions.

#11 about 3 min

Evaluating video encoding formats for faster decoding

Choosing older encoding standards over high-compression formats prevents slow decoding from bottlenecking the compute pipeline.

#12 about 5 min

Measuring optimized pipeline results and shifting bottlenecks

Achieving sub-minute processing times per video hour exposes downstream stages like tracking as the new limiting factor.

#13 about 1 min

Key lessons for large-scale machine learning performance

Practical guidelines for using concurrency and profiling to build reliable, high-throughput machine learning systems.

Matching moments

5:40 min

Deploying optimized models for high-throughput video analytics

Ekaterina Sirazitdinova · World Congress 2023

1:37 min

Optimizing latency and cost with free open source software

Rahima Yakoob · World Congress 2023

3:58 min

Controlling hardware cameras and analyzing visual frames

Marc Rousavy Marc Rousavy · World Congress 2025

51 sec

Overcoming inefficiencies in computer vision modeling

Antonio Tavera Antonio Tavera · World Congress 2025

1:44 min

Strategies for scaling AI workloads across multiple GPUs

Lerna Ekmekcioglu Lerna Ekmekcioglu · Europe 2026 Virtual

3:16 min

Composing real time video flow applications utilizing multiple AI models

Ankit Patel Ankit Patel · World Congress 2024

More from World Congress 2026 Europe - Virtual Stage

One Does Not Simply Make Architecture Decisions Alone thumbnail
36:15
2 months ago

One Does Not Simply Make Architecture Decisions Alone

Vladas Diržys

From Black Box to Glass Box : Bedrock AgentCore Observability thumbnail
21:15
3 months ago

From Black Box to Glass Box : Bedrock AgentCore Observability

Yasemin Aktürk

Leading in Rhythm: How Women Make Better Decisions with Biological Intelligence thumbnail
30:41
3 months ago

Leading in Rhythm: How Women Make Better Decisions with Biological Intelligence

Julia Grosse

Better Together: Leveraging Your Observability Tools as a SIEM thumbnail
39:06
3 months ago

Better Together: Leveraging Your Observability Tools as a SIEM

Mathias Palmersheim

ChatGPT Apps: From Chat to Checkout thumbnail
30:29
3 months ago

ChatGPT Apps: From Chat to Checkout

Aaron Koivunen

Interior Decorating (TypeScript Decorators) thumbnail
29:22
3 months ago

Interior Decorating (TypeScript Decorators)

Michael Dowden

Related videos

30 Golden Rules of Deep Learning Performance thumbnail
46:30
LIVE • over 6 years ago

30 Golden Rules of Deep Learning Performance

Anirudh Koul

WWC24 - Ankit Patel - Unlocking the Future Breakthrough Application Performance and Capabilities with NVIDIA thumbnail
22:07
World Congress 2024 • about 2 years ago

WWC24 - Ankit Patel - Unlocking the Future Breakthrough Application Performance and Capabilities with NVIDIA

Ankit Patel

Hacking AI at the Edge of the Indian Ocean thumbnail
29:03
World Congress 2026 Europe • 3 months ago

Hacking AI at the Edge of the Indian Ocean

Anastasiia Stefanska, James Cha-Earley

Accelerating Python on GPUs thumbnail
59:43
LIVE • over 2 years ago

Accelerating Python on GPUs

Paul Graham

The state of MLOps - machine learning in production at enterprise scale thumbnail
47:19
LIVE • over 4 years ago

The state of MLOps - machine learning in production at enterprise scale

Bas Geerdink

AI Factories at Scale thumbnail
24:26
World Congress 2024 • about 2 years ago

AI Factories at Scale

Thomas Schmidt

Related articles

From learning to earning

Jobs that call for the skills explored in this talk.