OmniScope F4 Launches with AI-Assisted Defect Recognition at NDT 2026

Live from Nashville: The First AI-Native Industrial Videoscope

Sugia Vision Technology chose the NDT 2026 conference in Nashville, Tennessee — the premier gathering of nondestructive testing professionals in North America — to publicly debut the OmniScope F4, a modular multi-probe videoscope platform with integrated AI-assisted defect recognition (AIDR). The F4 is the capstone of the OmniScope product line and the first industrial borescope to ship with a dedicated neural processing unit (NPU) on board.

The AIDR engine runs on a 24 TOPS INT8 inference accelerator within the console and has been trained on a curated dataset of over 250,000 NDT inspection images annotated by ASNT Level III certified inspectors across aerospace, power generation, and oil and gas domains. The model detects six defect classes — corrosion, cracking, coating loss, foreign object debris (FOD), erosion, and pitting — with a per-frame inference latency under five seconds. The system operates in “assist” mode: it highlights candidate anomalies with bounding boxes and confidence scores, queues them for operator review, and requires explicit operator confirmation or rejection of every detection. Rejected detections feed back into the model training pipeline under an optional fleet-learning program.

Why Operator-in-the-Loop Matters

“We explicitly chose an assist architecture rather than a replace architecture,” explains Dr. Wei Zhang, Sugia’s Director of AI Systems. “NDT inspectors carry legal and regulatory accountability for every accept/reject decision. The AIDR engine is a productivity multiplier — it reduces the cognitive load of scanning for anomalies in long inspection sequences — but the inspector remains the authority.” In testing with a major European airline’s engine MRO facility, the AIDR engine reduced per-engine inspection documentation time from 108 minutes to 64 minutes (a 41% reduction) while improving inter-inspector consistency by 22% as measured by Cohen’s kappa coefficient across five inspectors examining the same turbine component set.

The F4 also ships with an automated report generator that compiles inspection metadata, operator-confirmed findings with measurements and images, and a configurable severity table into ASME, API, and EN-compliant PDF reports. Native connectors for SAP PM, IBM Maximo, and Infor EAM feed inspection data directly into enterprise asset management workflows. The console manages up to 5,000 asset records with full inspection history and next-due-date tracking, providing fleet-level condition trending dashboards.

The OmniScope F4 will begin shipping in Q3 2026. Visit the Sugia booth at NDT 2026 (Hall B, Booth 427) for a live demonstration, or contact our engineering team to schedule a virtual briefing.