Release 2026/02/25

v3.8.4

Observ Feb Updates

This release focuses on a more user-friendly entry experience, privacy protection (face & license ㄙanonymization), richer event lifecycle control (start/end conditions), and key upgrades across VLM Flow, analytics, and model quality—helping you understand what’s happening faster and operationalize detections more easily.

New Features:

New Homepage (Faster, friendlier entry)

The redesigned homepage brings the most important info into one place:

  • Ongoing events

  • Task status

  • Live Feed

  • Quick access to analytics

  • Docs and resources

Face & License Anonymization

Observ now automatically blurs detected faces and license plates. This ensures stored event images, snapshots, and related assets are anonymized by default—supporting privacy protection and real-world compliance needs.

Event Start/End Conditions (BBox Custom Events)

BBox custom events now support an On/Off start & end mechanism. You can define separate rules for when an event starts and when it ends—enabling more flexible event lifecycles beyond one-time triggers.

New In/Out Analytics Charts

Zone-based In/Out now includes analytics charts for:

  • Hourly ingress / egress

  • Hourly net flow

Great for spotting trends, peaks, and anomalies.

15 New VLM Preset Scenarios

To reduce prompt setup friction, Create Task now includes 15 new VLM preset scenarios, helping users get started quickly without writing prompts from scratch.

  • Trespassing Event Analysis

  • High Volume Crowd Analysis

  • Traffic Accident Event Analysis

  • Post Natural Disaster Damage Analysis

  • Traffic Monitoring and Analysis

  • Stray Animal Monitoring and Analysis

  • No Smoking Violation Analysis

  • Fire and Smoke Event Analysis

  • Pedestrian Walkway Risk Detection

  • Queue Density Analysis

  • Air Quality Analysis

  • Firearm Event Analysis

  • Driving Surface Condition Analysis

  • Flood Analysis

  • Person On The Ground Analysis

✅ General Model Update (V5.0)

Based on recent detection error patterns, we improved model quality by:

  • Fixing annotation consistency for human posture-related labels

  • Adding more coverage and training data for ground pothole cases


Coming Soon:

We’re preparing the next wave of updates to improve configurability, integration, and flexibility in VLM-based decisioning:

  • API V2 (Richer Event Type configuration + better integration) We’re rolling out API V2 with pagination, more complete Event Type configuration details (pipeline, conditions, parameters, etc.), along with stronger batch capabilities and additional/updated endpoints—making integrations and data pipelines easier and more consistent.

  • Multi-logic evaluation for VLM outputs VLM results will support more advanced decision logic—such as multiple condition groups (AND/OR), layered rules, and flexible trigger combinations—so event definitions can be more precise and adaptable.

  • Event judgment for single / batch images A new feature is coming to run event detection on single images or batch image sets, ideal for playback, validation, and phone image uploading applications.

Stay tuned—these upgrades are just around the corner and will further enhance the scalability and efficiency of Observ in real-world deployments.


Support

For technical support, please contact our support teamarrow-up-right during business hours.

Last updated