Lexset

Founded by Amir Kashani in 2018

Turn live sensor feeds into geolocated inspection reports and generate synthetic computer vision datasets

Pricing model

Paid

Rating

★ People love it

Time to value

Quick Setup (< 1 hour)

You can use Lexset to turn raw sensor feeds into geolocated reports and generate synthetic training data. Its Sentinel system ingests full-motion video from drones or surveillance cameras, tracks objects across frames, calculates exact coordinates, and drafts standard reports before operations finish. You can deploy it on edge devices, servers, or cloud environments. Meanwhile, its Seahaven engine generates photorealistic, labeled 3D synthetic datasets to train computer vision models on rare scenarios, uncommon weather conditions, sensor noise, and specific structural defects without manual annotation.

What Lexset does

Process raw video feeds into structured incident reportsGeolocate detected objects to precise latitude and longitudeGenerate photorealistic 3D synthetic training imagesExtract object bounding boxes and depth maps automaticallyInspect infrastructure assets against engineering standardsExport training datasets directly into PyTorch and TensorFlowIngest MISB-compliant telemetry metadata alongside videoAnalyzes drone and sensor feeds in real timeCalculates exact coordinates for every visual detectionDrafts standard operational reports automatically in under sixty secondsRuns on edge hardware including NVIDIA Jetson devicesGenerates fully annotated synthetic 3D training data on demandProduces pixel-perfect semantic segmentation, depth maps, and bounding boxesIngests standard MISB-compliant KLV video feeds across infrared and radar