AI Inspection Atlas · summary as of November 2, 2026
AI Inspection Atlas Summary: 46 published deployments at a glance
Every number on this page is counted from the Atlas data, and each result is quoted from its public source. Vendor-published results are vendor claims and were not independently verified.
Download all 46 cases (CSV) Get new cases every Monday Print / save as PDF
By industry
- Automotive: 34
- Aerospace: 12
By context
- Production: 35
- MRO: 6
- R&D project: 5
By defect category (a case can have several)
- Adhesive & sealant: 5
- Assembly / missing part: 11
- Casting & machining: 6
- Composite structures & layup: 2
- Cracks: 5
- Engine blade damage: 4
- Labels & OCR: 6
- Other: 2
- Paint & coating: 7
- Surface scratches/dents: 17
- Weld & spatter: 9
What the sources disclose
- Camera / lighting stated: 39 of 46
- Training data described: 15 of 46
- Vendor case studies: 18 of 46 (the rest are OEM releases, trade press, integrator and research sources)
Vendors / developers with 2+ cases
- UnitX: 10
- SwitchOn: 4
- BMW Group: 3
- Waygate: 2
- Nissan: 2
- MVTec: 2
- Cognex: 2
All 46 cases
| Company / plant | Part · defect | Vendor | Reported result (quoted) |
|---|---|---|---|
| GE Aerospace MRO network (GEnx, CFM LEAP) Aerospace · MRO |
High Pressure Compressor (HPC) blades Engine blade damage (assisted defect recognition) |
Waygate Technologies (Baker Hughes) + GE Aerospace | "help increase detection rates by ~34%, while reducing false alerts by >13%" source |
| GE Aerospace / GEnx-1B and -2B operators Aerospace · MRO |
HPT stage 1 & 2 blades Blade damage screening and dispositioning |
Waygate Technologies + GE Aerospace | Not quantified; aims to "reduce operator variability" source |
| Pratt & Whitney (GTF, V2500 engines) Aerospace · MRO |
Aircraft engine Engine part condition/availability (lease inspections) |
Awiros (Percept, with Pratt & Whitney) | "Percept automates this inspection, and reduces time taken by nearly 90%" source |
| Boeing Commercial Airplanes, Renton, WA (737) Aerospace · Production |
Installed aircraft parts (part tags) Wrong/unverified installed parts (part/serial numbers) |
Boeing (in-house, Boeing Korea + Boeing AI) | "improve inspection time by more than 17 hours per airplane" source |
| Boeing solar panel manufacturing (MxD project) Aerospace · R&D project |
Space-grade solar panels (glue dots) Adhesive dot defects (tails, connections, position, shape) |
MxD project team (with Boeing) | Project goal; states it "replaces 14 human workers currently needed in this operation" source |
| Spirit AeroSystems, Wichita, KS (planned Prestwick) Aerospace · Production |
Fuselages, wings, substructures Composite structure defects (NDI) |
Spirit AeroSystems (in-house) | "sometimes up to 40 percent faster"; analysis "can be automated through machine learning" source |
| Airbus Atlantic (with IRT Jules Verne) Aerospace · R&D project |
Air intake acoustic panels Cosmetic defects: scratches, dents, surface defects |
IRT Jules Verne | "the solution has been successfully tested on real parts, under industrial conditions"; production integration planned "within two years" source |
| ZAero consortium (Profactor, FIDAMC and others) Aerospace · R&D project |
Wingskins / large CFRP parts Automated fiber placement layup defects |
ZAero consortium (Profactor et al.) | Project goals: "15% higher production rate, 15-20% lower production cost and 50% less waste" (goals, not measured results) source |
| ST Engineering Commercial Aerospace, Changi Creek airframe MRO facility, Singapore Aerospace · MRO |
Airframe upper surfaces (drone) and landing gear (robotic-arm camera) Corrosion, paint defects and surface dents |
ST Engineering (in-house AI-VET robotic-arm camera; drones) | Landing-gear camera on a robotic arm is "reducing inspection time by up to 83 per cent"; "Artificial intelligence models help to identify corrosion, paint defects and surface dents." source |
| Safran (cabin / lavatory parts); rollout to Marysville, WA and Santa Maria, CA Aerospace · Production |
Aircraft lavatory parts (toilet lids/seats) and cabin assemblies Cosmetic paint defects: scratches, paint discoloration, paint bubbles |
Loopr AI | Toilet-seat inspection went from "20 to 30 minutes" to "five to 10 minutes"; "90.9% recall"; "we're able to produce 10% to 15% more than before" source |
| Pratt & Whitney (RTX) MRO network incl. GTF MRO network shops; used by lessors and MROs on V2500 engines Aerospace · MRO |
Aircraft engine parts inspected by borescope (V2500; pilots on GTF and F135) Engine defects recognized and recorded from borescope video (AI-assisted defect recognition) |
Aiir Innovations (acquired by Pratt & Whitney, July 2026) | "lessors have recorded instances of up to 75% faster analysis of borescope videos, while for MROs, the software has reduced overall inspection time by more than 30%, including a reduction in report-generation time from hours to minutes" source |
| Jet Aviation MRO, Basel (Switzerland); non-regulated inspections in the US (managed and FBO customers, e.g. Burbank demo) Aerospace · MRO |
Aircraft exterior surfaces (business jets, narrow-body and some wide-body types) Aircraft exterior surface condition: general visual inspection findings, paint quality |
Donecle (drone and AI technology), with Jet Aviation | No inspection metrics published. Images "are then analyzed using specialized software to identify potential areas of interest. Maintenance professionals review and validate the findings" before a digital report is compiled. Approved by Swiss FOCA in 2024 for General Visual Inspections (GVIs) by images; US use (since May 2026) is for non-regulated inspections. source |
| Unnamed "leading Japanese automotive brand" (US factories) Automotive · Production |
Part not stated Production-line defects (unspecified) |
FPT Software + LandingAI | "achieving over 99% accuracy within a scanning time of just 3–5 seconds" source |
| Unnamed engine plant (integrator: Automation NTH) Automotive · Production |
Engine rocker arms Assembly verification (clip/rocker arm labeling) |
Cognex (VisionPro ViDi) via Automation NTH | "The ViDi system now achieves 98% to 99% accuracy in labeling inspected rocker arms. This allowed the customer to eliminate a manual inspection station from the line" source |
| Unnamed supplier to an automotive OEM Automotive · Production |
Transfer case housing (sealing surface) Casting/machining: non-cleanup, flashing, chunking |
UnitX | "False Acceptance Rate: 0% . False Rejection Rate: 0%"; "Days to SAT: 5 Days"; "Cycle Time: 0.39 seconds" source |
| Unnamed Automotive · Production |
Connecting rod Scratches, dirt, press marks, air bubbles, short welds, cracks, whitening (exposed substrate) |
UnitX | "False Acceptance Rate (FA): ≤ 1%"; "Cycle Time: ≤ 15 seconds per piece" source |
| Unnamed Automotive · Production |
Pump housing (~60x60x70mm) Porosity, scratches, pits, dents, color variation, peeling, incomplete machining |
UnitX | "0% False Acceptance Rate for safety-critical defects"; "False Acceptance of ≤ 0.5% for cosmetic flaws and a False Rejection Rate of ≤ 5%"; "5-second cycle time" source |
| Unnamed Automotive · Production |
Automotive connector Scratches, blowholes, flash, exposed copper, weld misalignment, glue overflow |
UnitX | "0% False Acceptance Rate"; "≤ 10% False Rejection Rate"; "≤ 7s Cycle Time" source |
| Unnamed Automotive · Production |
ADAS camera module / vehicle lens assembly Optical contamination, adhesive bubbles/missing glue, slanted locking |
UnitX | "Achieved a 100% catch rate for critical defects like aperture contamination and oil stains"; "1.5 seconds per unit"; "False Rejection Rate ≤ 10%" source |
| Unnamed Automotive · Production |
Zinc die-cast parts Blisters, incomplete filling, burrs, scratches |
UnitX | "0% False Acceptance Rate"; "≤ 5% False Rejection Rate"; "0.72 s/pc Cycle Time"; "5,000 pcs/hour Capacity" source |
| Unnamed Automotive · Production |
Automotive sleeve Dents, scratches, weld slag, porosity, burrs, thread/assembly errors |
UnitX | "0% False Acceptance Rate (Critical Defects)"; "< 5% False Rejection Rate (Overall)"; "≤ 4.5s Cycle Time"; "≥ 95% Overall Equipment Effectiveness (OEE)" source |
| Unnamed OEM Automotive · Production |
EV transmission motor shaft Rust, oil stains, dents, scratches, haze, missing holes |
UnitX | "0% False Acceptance Rate (Level 1 Defects)"; "≤ 5% False Rejection Rate"; "≤ 9 s Cycle Time" source |
| Unnamed Automotive · Production |
Three-way valve flange Rust, scratches, dents, burrs, tool/clamping marks |
UnitX | "0% False Acceptance Rate"; "≤ 5% False Rejection Rate"; "≤ 26 s Cycle Time"; "≥ 90% Overall Equipment Effectiveness (OEE)" source |
| Unnamed Automotive · Production |
Chassis rubber bushing Adhesion failure, cracks, indentation, foreign objects |
UnitX | "≤ 0.1% False Acceptance Rate"; "≤ 5% False Rejection Rate"; "≤ 9s Cycle Time" source |
| BMW Group, Dingolfing press shop Automotive · Production |
Formed sheet-metal body parts Fine cracks vs. pseudo-defects (dust, oil) |
BMW Group (in-house) | "With the new AI application, these pseudo-defects no longer occur" source |
| BMW Group, Dingolfing final inspection Automotive · Production |
Vehicle model badges (e.g. "xDrive") Missing/wrong model designation badges |
BMW Group (in-house) | "If the live image and order data don’t correspond, for example if a designation is missing, the final inspection team receives a notification." source |
| Audi, Neckarsulm (rollout to Audi Brussels, VW Emden, Ingolstadt) Automotive · Production |
Car body (body-in-white) Resistance spot-weld quality anomalies |
Audi (in-house, WPS-Analytics) | "Audi analyzes around 1.5 million spot welds on 300 vehicles each shift at its Neckarsulm site." source |
| Audi body shop Automotive · Production |
Car body underbody Weld spatter |
Audi in-house model on Siemens Industrial Edge / Industrial AI Suite | "The AI reliably detects weld splatter and helps workers remove it" source |
| Porsche, Leipzig Automotive · Production |
Vehicle labels (airbag notes, vehicle info) Wrong label content or language |
Volkswagen Group / Porsche (in-house) | Not stated (group expects "cost reductions running into the double digit million euro range" from computer vision overall) source |
| Unnamed "leading Indian automotive multinational" Automotive · Production |
Crankcase Sealant application defects |
SwitchOn (DeepInspect) | "Achieved 99.5% accuracy" (summary also says "over 95%") source |
| Unnamed auto component manufacturer Automotive · Production |
Connecting rod Surface defects |
SwitchOn (DeepInspect) | "Detected surface defects up to 150 microns with 99.5% accuracy"; "The false positive rate was minimized to just 0.5%" source |
| Unnamed large Indian automaker Automotive · Production |
Injector clip Missing/misfitted clips |
SwitchOn (DeepInspect) | "the system guaranteed 100% inspection success rate"; "The false positive rate is reduced to less than 2%" source |
| Unnamed global automotive component supplier Automotive · Production |
Starter switch Caulking defects |
SwitchOn (DeepInspect) | "Defects as small as 100-150 microns were able to detect"; "Inspection completed in less than 11 seconds per part" source |
| DGH Group for a large French automotive group Automotive · Production |
Body-in-white MIG and laser welds Cracks, incomplete weld seams, irregular welding patterns |
MVTec HALCON + Deep Learning Tool (integrator DGH Group) | Page says detection rates are "very high and robust" (quoted from DGH); no figure stated source |
| Panasonic Energy, Kansas plant (US) Automotive · Production |
Electrode tabs of cylindrical Li-ion cells (EV batteries) Fiber waste adhesion, welding spatter on laser weld marks |
MVTec HALCON | "resulted in a 57.3% reduction in the number of over-detected discharges"; headline: "Zero defective outflow achieved while halving the number of over-detections" source |
| Toyota Motor Manufacturing UK Automotive · Production |
Vehicle body (in-line) Vehicle build checks (60+ inspections per vehicle) |
Acrovision using Cognex VisionPro + ViDi | "significant reductions in inspection times" (first phase) source |
| Nissan Smyrna, TN; also Canton, MS and Aguascalientes, MX Automotive · Production |
Painted vehicle body Paint flaws: dirt, fibers, craters |
Nissan (AUTIS system) | "The human eye can spot 85%-95% of flaws, but AUTIS identifies over 98%"; finds "imperfections as tiny as 0.2 millimeters" source |
| Ford, 17 North American plants incl. Kentucky Truck Plant Automotive · Production |
Vehicle assembly (connectors, wiring, door panels) Missed components, misaligned assembly, unseated connectors, wrong parts |
IBM Maximo Visual Inspection | "150 million individual inspections ... that flagged 400,000 quality issues that a human worker may have missed" source |
| Yamaha Motor, ATV/ROV division Automotive · Production |
ATV/ROV warning labels Wrong/misplaced warning labels |
Elementary | "improved the speed and accuracy of label inspections while providing complete traceability" source |
| Hyundai Motor and Kia (E-FOREST Center, Uiwang R&D; planned for Kia Hwaseong EVO Plant West) Automotive · R&D project |
Finished vehicle exterior (Kia PV5 shown in demo) Exterior scratches and other surface abnormalities on finished vehicles |
Hyundai Motor and Kia (in-house, E-FOREST Center Vision Task Force) | "Each inspection involves capturing and analyzing about 5,000 ultra-high-resolution images to check for scratches and other exterior abnormalities, with the location and size of any defect displayed on screen." No detection-rate or accuracy figures given. source |
| BMW Group, Shenyang high-voltage battery plant, China Automotive · Production |
Gen6 high-voltage battery (cylindrical cells, cell-to-pack) Deviations in laser-welded joints (cell clusters to cell contact system) |
BMW Group (in-house) | Not quantified. BMW: "Digital twins, data analytics and AI-supported inspection methods help detect deviations at an early stage and ensure traceability. Among other parameters, inline inspection systems monitor laser-welded joints in real time." source |
| AUTIS (industrial monitoring and automation systems) with TECNALIA; validated on images from six vehicle manufacturers, eight plants Automotive · R&D project |
Painted vehicle bodies (car bodywork) Paint surface defects (AI generates the inspection masks that define paint-inspection zones) |
AUTIS with TECNALIA | "automatic mask generation via computer vision and Artificial Intelligence is technically feasible and reduces the effort required to set up new inspection systems" source |
| Hyundai Motor India, Pune plant (Maharashtra, India; former GM Talegaon), press shop Automotive · Production |
Stamped steel body panels (5,400-ton press line, all-new Hyundai VENUE) Fine cracks and surface defects on stamped panels |
Hyundai Motor Company (system supplier not disclosed) | "an AI-based panel crack detection system shares that work. Using infrared (IR) sensors together with camera vision inspection, it automatically catches even fine cracks and surface defects" source |
| Nissan Industrial Complex, Resende (Rio de Janeiro, Brazil); Kicks and Kait production; solutions also adopted at Nissan plants in Mexico and the US Automotive · Production |
Vehicle body (carroceria); installed wheels and tires Bodywork defects; wheels and tires that do not match the vehicle's factory configuration (wrong part) |
Nissan (in-house; developed partly by Nissan Brazil teams) | No inspection metrics published. Nissan says the plant has 30 AI initiatives, "dos quais 20 já estão em operação" (20 already in operation), including "robôs equipados com câmeras capazes de realizar inspeções automatizadas da carroceria" and tools that verify wheels and tires match "a configuração correta de fábrica". source |
| Hyundai Motor and Kia production operations in Korea (incl. Ulsan, Asan, Jeonju, Gwangmyeong, Hwaseong, Gwangju), the United States, Europe, India and Asia-Pacific Automotive · Production |
VIN (vehicle serial number) read from vehicles on the production line Vehicle/specification mismatch: physical vehicle not matching registered system data (assembly-sequence errors, risk of wrong part) |
Hyundai Motor Group / Kia (in-house Vision AI team; inference on Amazon Web Services) | Hyundai Motor Group says the AI Automation Recognition Service is "deployed across approximately 70 manufacturing processes" and is "generating annual cost savings of approximately KRW 5.24 billion (approximately $3.9 million)." No detection-accuracy figures published. source |
| General Motors Factory ZERO (Detroit-Hamtramck, Michigan, USA) Automotive · Production |
Body-in-white resistance spot welds Resistance spot weld quality |
General Motors (WeldBrAIn) | GM says WeldBrAIn "assesses every resistance spot weld on every vehicle in real time." Per the article, it "expanded quality check capabilities from manual inspection of four parts per shift to checking every single weld, on every single body, in real time." No defect-rate or accuracy figures published. source |
Spot an error or know a public case we missed? Tell us.