AI4RVI has launched Edgevision, a remote visual inspection (RVI) system that pipes true 1920×1080 HDTV video through a 176-degree wide-angle lens and processes everything onboard an NVIDIA Jetson computer — no network connection required.

The Jetson module handles video recording, image capture, and all camera functions locally. Because processing happens on the device itself, Edgevision runs 100% offline. That design choice addresses a real constraint of industrial inspection sites: boiler houses and heat-exchanger galleries often sit in radio-hostile or network-restricted areas where cloud-dependent inspection tools stall.

What can the system physically reach?

The camera head measures under 1 inch (25 mm) in diameter, detaches, and is submersible. It connects through a 98.4-foot (30 m) pushable cable, which lets inspectors image tube interiors that conventional borescopes struggle to reach clearly. AI4RVI optimized the optics specifically for boiler and heat exchanger tube ID inspection, where the 176-degree field of view covers the full inner wall circumference as the probe advances.

The complete system weighs under 20 lbs. (9 kg), a figure that matters for a specific access problem: the package fits through steam- or mud-drum manways, the narrow personnel openings that constrain what inspection gear can enter a drum at all.

How do the AI features work?

Edgevision offers two optional AI-powered capabilities, and buyers should note both are add-ons rather than baseline features:

• AI-assisted report generation, intended to accelerate post-inspection documentation.
• Auto-anomaly detection with standardized anomaly coding, which flags potential issues as the inspector works rather than after the fact.

The standardized coding aspect deserves attention from QA teams. Anomaly detection that maps to a consistent coding scheme — rather than free-form flags — is what makes findings comparable across inspectors, plants, and inspection campaigns. That comparability is the foundation of any trending program that tracks tube-wall condition over successive outages.

What comes in the package?

The hardware configuration includes several details aimed at field workflow:

• Wi-Fi control via a Lenovo Yoga Tab 11 with a 3.2K display hosting the Edgevision software
• Supervisory viewing over Wi-Fi plus HDMI video output
• External video input through a USB port for alternate sources, with a footage counter display
• Two hot-swappable batteries for continuous cordless operation
• An ATA shipping case, three camera centering fixtures, and a keyboard

The hot-swappable batteries remove the mid-inspection shutdown: a tube campaign interrupted by a dead battery means re-inserting the probe and re-covering ground. The three centering fixtures keep the camera concentric within different tube IDs, which directly affects image sharpness and anomaly localization on the wall.

Measured performance versus claims

One caution for evaluators: AI4RVI specifies the video pipeline as true 1080p HDTV through the full chain — capture, processing, and display — rather than a sensor upsampled to HD output, a distinction that separates this class of system from cheaper scopes advertising interpolated resolution. The company has not published accuracy figures for the AI anomaly detection, its false-positive rate, or the specific coding standard the anomaly codes follow. Procurement teams running qualification trials should request detection-performance data under representative tube conditions before relying on the automated flagging, and treat the AI output as a triage layer over inspector judgment rather than a replacement for it.

"With Edgevision, we set out to give inspectors a system that finally matches the true HDTV imaging the industry has been asking for, without sacrificing the portability and offline reliability that fieldwork demands," said Bruce Pellegrino, VP Marketing, AI4RVI, Inc.

The adoption question

The open question is how inspection organizations will fold onboard AI into qualified procedures. Insurance-backed boiler inspection regimes and heat-exchanger programs under API or ASME frameworks typically certify the inspector and the method — not the algorithm. Until anomaly-coding schemes from edge-AI systems map cleanly onto established reporting standards, plants will likely run Edgevision's automation in parallel with conventional callouts, using it to speed reporting while human reviewers retain sign-off. How quickly AI4RVI documents that mapping may determine whether the Jetson inside is a convenience or a workflow change.