OIL & GAS
Methane Leak & Emissions Monitoring
Methane Leak & Emissions Monitoring. What it takes to run it.
Visual detection of gas plumes via infrared/thermal imagery.
// 01 · The question
“Is that a plume, and which piece of equipment is it coming from?”
// 02 · Why it is hard
The signal is a subtle contrast change in infrared, not an object. Steam, heat shimmer and exhaust all look like the thing you are hunting, and the difference is often behaviour over several frames.
// 03 · What the labels have to be
Instance segmentation
Decided before the first frame
A plume is a soft, moving, semi-transparent region. It has no box worth drawing, and both the questions asked of it, how large and where it originates, are read off the shape.
This is the decision that is expensive to reverse. The geometry has to match what the answer contains, and finding out it does not means labelling the set a second time.
// 04 · How we run it
Three parts of one loop, on this job.
LexAnnotate
Precision Safety
Annotate rig images with bounding boxes for machinery, valves, flanges, and personnel with PPE classification. Safety-focused labels tailored to oil & gas operations. 12k+ hazardous-zone annotations in production deployments.
LexInsight
Corrosion Tracker
Link visual detections to asset management systems. Map corrosion to specific valve IDs, track progression across inspection cycles, connect to maintenance scheduling and compliance codes.
LexAlert
Safety Alerts
Monitor deployed safety models against environmental changes - weather, lighting, equipment repositioning. Detect drift before it causes a missed hazard. Route failures back for immediate retraining.
// 05 · What breaks it after launch
Optical gas imaging cameras are calibrated instruments. A replacement unit with a different sensitivity or palette changes what a plume looks like at the pixel level even though nothing at the site has changed.
It is not the only one that can get this use case, it is the one that usually gets it first. All ten conditions.
// 06 · In oil & gas
Hazardous Zone Monitoring
The Challenge
Lacked labeled rig imagery, delaying AI model training. Relied on manual monitoring in hazardous zones.
Our Execution
Annotated rig images with bounding boxes for machinery and personnel. Delivered precise, safety-focused labels.
The Result
Faster AI deployment with rapid model iteration. Improved safety through automated detection.
12k+
Precision image annotations delivered