Improved: Open bypass switch detection

Graduating from beta, this model now detects open bypass switches with 79% accuracy. This model is particularly helpful in storm patrol use cases to enable faster identification of downed circuits and streamline power restoration workflows.

Improved: Insulator composition detection

Updated model performance for distinguishing between polymer and porcelain insulators, achieving 80% accuracy across cutouts, strain insulators, pin insulators, and arresters. This upgrade supports more precise asset condition assessments for utilities looking to replace their older porcelain insulator components.

New: Day burner detection

Previously in beta, our day burner detection model is now available to all customers. This model identifies streetlights with faulty photocells that remain on during the day, with an accuracy rate of 87%. This feature helps support cost savings by improving energy efficiency.

New: NEMA socket detection

Also transitioned out of beta from last quarter, this model detects the presence of NEMA sockets on streetlights with ~90% accuracy, supporting IoT hardware upgrades and enabling smart city solutions through compatible device mounting.

New: Dashboards

Quickly surface high-level insights during deployments to simplify reporting and surface trends across different asset types, issues, regions and more.

New: Noteworthy insights

Instantly visualize critical infrastructure issues in collected imagery. Top-level buttons highlight AI-detected problems such as corroded or leaking transformers, severely leaning poles, and porcelain insulators, bringing urgent issues to the forefront.

New: Lighting asset summary

A new lighting section in the asset details panel displays the fixture type, light source (LED vs. non-LED) and other important lighting asset information directly in the asset details pane.

New: Edit customer asset matches

Users can now correct mismatches between Noteworthy-detected assets and customer-provided GIS data. If an asset is incorrectly matched (or not matched at all) to the customer-provided asset ID, the association can be updated directly in the platform—improving overall data accuracy and consistency.

New: Edit asset type

Users can now manually update the type of an asset (e.g., utility pole vs. light pole) in Inspect Cloud. This provides greater flexibility when reviewing AI results and helps maintain clean, accurate asset inventories.

Improved: GPS connection monitoring

New GPS monitoring logic checks for dropped connections and attempts to automatically reconnect, ensuring more consistent vehicle GPS and fewer geolocation gaps.

Improved: Frame-saving logic for edge cases

Updated logic ensures that valid images—previously discarded due to overly strict filtering—are now retained, increasing asset coverage and improving overall data quality.

Improved: Vehicle auto-shutdown handling

For vehicles with auto-shutdown enabled while idling, the system now uses GPS-based movement detection to prevent premature stack shutdowns caused by false ignition readings. This ensures consistent uptime and data collection.