Skip to main content

Blue Iris 6 – AI Tuning

🔹 Overview​

This page captures practical guidance for using AI confirmation in Blue Iris to improve event quality and reduce false alerts.

The focus is on disciplined setup: clean motion first, AI second, alerts third.

🎯 Objective​

  • Improve event quality
  • Reduce false alerts
  • Build a stable AI confirmation workflow
  • Avoid using AI to compensate for poor trigger design

🧠 Key Concepts​

  • AI confirms events; it should not replace sensible motion settings
  • Image quality matters to AI results
  • Confidence thresholds should be tuned, not guessed
  • Stable workflows beat aggressive settings

🛠️ Core Tuning Areas​

Motion Before AI​

  • Start with reasonable motion tuning first
  • Reduce obvious false triggers from scene noise
  • Confirm the camera is generating usable trigger events

Image Source​

  • Make sure AI is analyzing a useful image
  • Poor framing, weak detail, or excessive compression reduces AI value
  • Confirm whether main stream or sub-stream gives the better practical result for your use case

Confidence Thresholds​

  • Start with moderate confidence settings
  • Review confirmed and unconfirmed events
  • Raise or lower confidence only after observing actual results

Object Classes​

  • Enable only the classes you truly care about
  • Extra categories can create noise and unnecessary reviews
  • Keep the workflow aligned with the actual security objective

Alert Logic​

  • Do not enable aggressive notifications immediately
  • First confirm that AI-confirmed events are consistently meaningful
  • Then layer on notifications and external actions

Performance​

  • AI adds processing load
  • Watch CPU, GPU, and overall responsiveness together
  • Scale carefully when enabling AI across many cameras

📊 Recommended Baseline​

  • Reasonably clean motion
  • Moderate confidence threshold
  • Limited object classes
  • Simple alert logic
  • Review actual results before expanding

🧠 Real-World Notes​

  • Most “AI problems” start as motion or scene-design problems
  • Good framing usually improves AI more than aggressive settings
  • Too many object classes often create more review work than value
  • A slower, cleaner rollout usually produces better long-term results

⚠️ Common Mistakes​

  • Turning on AI before motion is usable
  • Letting AI analyze poor images
  • Enabling too many classes
  • Sending notifications before validation
  • Changing multiple AI variables at once

📊 Related Pages​

✅ Result​

This page provides a practical framework for using AI confirmation in Blue Iris without creating extra noise or instability.