We generate missing data, and find hidden risks
IPIS Lab synthesises AI training data for rare and hazardous conditions that cannot be captured in the field (DataSynOps), and uses that data to read infrastructure imagery automatically (InfraSafe AI).
2 are repeat buyers
incl. K-water open innovation
6 verified in delivery
Defense · Infrastructure · Environment · Manufacturing
1 filed · 2 software regs.
What blocks an AI project is the data, not the model
Models can be downloaded. The training data for your site cannot be bought anywhere.
Rare events, by definition, leave no data
Voids, sinkholes, wildfires, industrial accidents — the very conditions a model must detect are the ones that almost never occur. Field capture alone will never reach training scale.
Collection and labeling consume the budget
Capture, cleaning and manual annotation absorb most of a project's cost and schedule, and the bill arrives again with every new site.
Security and privacy close the door
Defense, medical and site CCTV footage often cannot leave the premises at all, which rules out outsourced annotation before the work begins.
Human reading is not repeatable
Two inspectors mark the same crack differently, and every agency keeps its own criteria. Re-inspection reproduces the variance instead of removing it.
Generate exactly the data you need
Physics-based simulation and generative AI synthesise rare conditions at scale, with ground-truth labels attached automatically.
Diagnose the gap
We identify the conditions the current model fails on, automatically.
Synthesise the data
Physics-based simulation combined with generative realism refinement.
Auto-label
SAM2 produces pixel-precise polygons, sorted into three confidence tiers.
Retrain and verify
Synthetic pre-training, field fine-tuning, then measured verification.
Deploy and report
Detections and the evidence behind them are exported as a report.
We read the site with the data we generated
The engine stays the same when the domain changes. That is why entering a new market costs us little.
SewerDefect AI
Sewer networksReads defects from in-pipe CCTV footage and drafts the inspection report.
GroundSafe AI
Subsurface safetyDetects voids, leaks and ground loosening in GPR survey data and ranks drilling priority.
WorkerSafe AI
Worker safetyDetects hazardous worker situations in real time in low-light and confined sites.
StructureCrack AI
Built structuresDetects cracks and surface defects in buildings, tunnels and bridges and grades severity.
Clients have already paid for this
Not PoC credits — three consecutive years of delivery confirmed by contracts and tax invoices.
Sewer defect auto-labeling
The AI draws the labels; people only check the ones worth checking
Defense drone AI video analytics
Aerial-view scarcity solved with a generation pipeline instead of flight hours
Wildfire and algal bloom detection data
The pipeline built in year one was re-purchased in year two
Secured as rights, verified by third parties
The core of our generation method is held as granted patents and verified by certified testing and awards.
Intellectual property
- Method and apparatus for generating training data for AI video analytics modelsKR 10-2687011
- Integrated image enhancement and training-data generation using a generative recurrent networkKR 10-2654017
- Image matting method and apparatusKR 10-2624296
Certification
- 2 certified test reportsIssued by KCL
- 2 software registrationsKorea Copyright Commission
- 4 government programsKRW 396M awarded to date
Awards
- 2026 H2 Innovative IT AwardJoongAng Ilbo
- 2025 S.Challenge IR Grand Final, Excellence AwardSeoul Regional SMEs Office
- Selected for Seoul Pavilion, CES 2025SBA
An execution team built on thirty years of imaging research
A founder-led R&D organisation combined with an advisory network of world-class imaging researchers.
Professor, Department of Imaging Science, Chung-Ang University. Ph.D. in Electrical Engineering, Northwestern University (1990). Over thirty years in image processing and computer vision, with numerous national R&D programs led as principal investigator for the Ministry of Science and ICT, the Ministry of Food and Drug Safety and the Ministry of Culture, Sports and Tourism. Founded IPIS Lab in 2023 to commercialise the laboratory's technology.
Ph.D. in Image Engineering, Chung-Ang University (2020). Leads generative AI and defect-detection pipelines
Object detection and tracking. Built the SAM2 auto-labeling pipeline
3D vision and image processing
EO-IR image analysis
Joseph Cummings Professor, Northwestern University
Emeritus Professor, University of Tennessee, Knoxville
CSO, President — Twelve Labs
Tell us what has to be detected on your site
Tell us what has to be detected on your site. We will first assess whether that data can be generated. If you have sample footage, we will prepare an auto-labeling demonstration with it.