INSIGHTS
Latest thinking.
Field notes on AI inspection, agents and physical AI — across electronics, semiconductor, automotive, EV battery, pharma, food, chemical and consumer manufacturing.
AI Vision Inspection for SMT Solder Joints: Solving Problems of Cold Solder, Bridging, and Insufficient Solder
This article introduces the pain points faced by a leading electronics/PCBA manufacturer in SMT solder joint inspection, and how DaoAI's AI vision solution effectively solves these problems and achieves remarkable results.
Read more
AI Vision Inspection Solution for Hidden Solder Joints under BGA/QFN Packages
This article introduces the pain points encountered by a leading electronics/PCBA manufacturer in the inspection process of hidden solder joints under BGA/QFN packages, and how DaoAI uses advanced technologies to solve these problems and achieve remarkable results.
Read more
AI Vision Technology Empowers Micro - Crack Detection of Passive Components like MLCC
This article introduces the application of AI vision in the micro - crack detection of passive components like MLCC, and demonstrates how DaoAI products solve industry problems and achieve remarkable results.
Read more
AI Vision Helps the Electronics/PCBA Industry Solve the Problem of False Alarms for Component Polarity Reversal
This article introduces the application of WeLinkirt's AI vision technology in the detection of component polarity reversal in the electronics/PCBA industry, solving the false alarm problem of traditional detection methods and improving the detection efficiency and quality.
Read more
AI Vision Inspection Solution for Missing Components and Incorrect Connector Installation in PCBA Assembly
This article presents an AI vision application case by WeLinkirt for detecting missing components and incorrect connector installation in PCBA assembly, covering user scenarios, pain points, technical principles, solutions, and quantifiable results.
Read more
Taming Post-Reflow AOI False Calls: An AI Second Gate That Cuts Re-Review
A leading consumer-electronics PCBA plant fought chronic AOI false calls that buried its re-review stations. DaoAI AI-AOI software, paired with APDT positive-sample learning, sat as a…
Read more
3D AI Inspection of Hidden BGA/QFN Joints: Voids, Bridges, Non-Wetting in One Pass
At a communications-module contract manufacturer, BGA/QFN balls hide beneath the package—2D optics can't see them and X-ray sampling is slow and costly. DaoAI's 3D AI-AOI hardware…
Read more
MLCC Micro-Cracks: Live on 1–20 Good Parts, Micron-Scale, Milliseconds per Part
At a passive-component maker, MLCC end-face micro-cracks are only microns wide—hard for classic vision to catch—and defect samples are nearly impossible to collect. DaoAI APDT…
Read more
Component Polarity False Calls: 164 Images, <1s Inference, 99% Accuracy
At an automotive-electronics PCBA plant, diode and electrolytic-cap polarity calls were plagued by false positives—legacy AOI read silk glare and body color shift as reversed. DaoAI…
Read more
PCBA Missing Parts & Connector Misloads: Full Inspection at High Takt
At an industrial-control-board PCBA plant, connectors, jumpers, and heatsinks were prone to omission, misload, and wrong orientation, and manual final inspection couldn't cover it all.…
Read more
Post-AOI Wafer Defect Auto-Classification: From Manual Review to AI-ADC
A wafer fab relied on manual AOI review with an overkill rate above 20%. With DaoAI AI-ADC, overkill dropped below 2%, review headcount fell 90%, and line throughput rose 30%.
Read more
Post-Wire-Bond Inspection: Defect Models Pushed to the AI-AOI Edge
A packaging plant relied on manual visual inspection for post-bond defects like sagging wires, with occasional escapes. DaoAI embedded the defect model into the AI-AOI edge device: 98%…
Read more
Rare Defects in Advanced Packaging: Few-Shot APDT Positive-Sample Learning
At an advanced packaging plant, new-process defect samples were too scarce for conventional supervised models. DaoAI combined APDT positive-sample learning with world-model…
Read more
Advanced Packaging Bump Inspection: 3D AI-AOI Catches Missing Bumps and Bridging at Micron Level
At an advanced packaging plant, micro-bump missing and bridging were hard to separate by 2D grayscale. DaoAI 3D AI-AOI judges bump topography from micron-level height data, sharply…
Read more
Lead Coplanarity and Dicing Chipping: 3D Vision Measures Height Per Lead
At a packaging plant, lead coplanarity and dicing chipping were hard to quantify in 2D. DaoAI 3D vision measures height per lead and checks chipping per edge, cutting both coplanarity…
Read more
100% Inline 3D Sealant Inspection: A Brain-Eye-Body Closed Loop Stops Every Break for a Tier-1 Supplier
A Tier-1 supplier's door-panel sealing line was plagued by breaks and skips, driving downstream water-leak rework. DaoAI 3D robotic vision delivers 100% inline inspection with…
Read more
Random Bin Picking: 6D Pose Drives 99%+ Pick Success on Featureless Parts
A parts plant long relied on workers hand-picking metal parts from bins; chaotic poses and featureless surfaces repeatedly blocked automation. DaoAI 3D robotic vision uses 6D pose…
Read more
Body-in-White Weld Seam Inspection via Deep Learning: No Hiding for Cracks or Lack of Fusion
A body-shop plant relied on manual visual checks of body-in-white weld seams, where cracks and lack of fusion slipped through and standards varied by inspector. DaoAI AI-AOI introduced…
Read more
X-ray Porosity and Inclusion Inspection for Aluminum Die-Castings: Line-Speed Judgment Under 2s per Image
A parts plant needed X-ray to detect internal porosity and inclusions in aluminum die-castings, but manual film reading was slow, subjective and fatigue-prone. DaoAI deep-learning…
Read more
0.3mm-Class Surface Defect Inspection for Stamped Sheet Metal: Escape Rate from 2.8% to 0.2%
A body-shop plant faced small, low-contrast scratches and dents on stamped sheet metal, with manual visual escape rates as high as 2.8%. DaoAI AI-AOI delivers 0.3mm-class surface…
Read more
Inline Full Inspection of Electrode Coating: 3D AI-AOI Securing Cell Consistency at the Source
A leading power-battery manufacturer deployed DaoAI 3D AI-AOI at the coating stage, fusing point cloud and image data to detect dark spots, scratches, exposed foil and material…
Read more
Tab-Welding Quality Gate: Stopping Burrs and Cold Welds Before Thermal Runaway
A cell manufacturer added a DaoAI AI-AOI automated quality gate to an existing tab-welding line, identifying burrs, misalignment and cold welds tied to thermal runaway, building 100%…
Read more
Cylindrical Cell Reflective Can-Surface Inspection: Taming Glare with Computational Imaging
A battery plant's cylindrical-cell steel cans are highly reflective with strong curvature; DaoAI's multi-light computational imaging suppresses glare, reaching judgment accuracy above…
Read more
Anode-Cathode Alignment in Winding and Stacking: 3D X-ray CT plus AI Layer Analysis at ~90x Speed
A leading power-battery manufacturer used DaoAI 3D X-ray CT with AI layer-by-layer analysis to replace manual reading, raising anode-cathode alignment and bend-inspection efficiency by…
Read more
Inline Inspection of Separator Pinholes and Coating Defects: Micron-Level Control on High-Speed Web
A battery plant's separator line runs fast and wide; DaoAI's high-speed web inline AI-AOI detects pinholes, coating scratches and thickness non-uniformity, achieving micron-level…
Read more
Visual Inspection of Sterile Injectables: How Deep Learning Draws the Line Between Particles and Bubbles
A sterile injectables plant was caught between missed particles and false bubble rejects. DaoAI AI-AOI re-inspection with APDT few-shot learning lifted particle catch by 70% and cut…
Read more
High-Speed Blister Inspection: Catching Every Missing Pill and Breach at 50,000 Units per Hour
A solid-dose plant's blister line ran fast, with glare and foil sealing defeating conventional vision. DaoAI AI-AOI detects missing pills, breakage, and seal defects in real time at…
Read more
In-Line Label OCR and Serialization: Letting Zero-Defect Labels Flow at Full Speed
A pharma packaging plant struggled with missing characters, misprinted barcodes, and serialization compliance. DaoAI OCR/OCV verifies characters, barcodes, and serialized codes in-line…
Read more
80 Classes of Capsule Defects: Operator Self-Training with Zero Code, Replacing Manual Inspection
A solid-dose plant faced up to 80 capsule defect classes with fatigue-prone manual inspection. DaoAI AI-AOI lets operators self-train with zero code and switch products in 5 minutes,…
Read more
Vials and Ampoules: Guarding Fill Level, Seal, and Glass Fragments Under High-Speed Inspection
After filling and sealing vials and ampoules, a sterile injectables plant must verify fill level, seal integrity, and clear particles at speed. DaoAI AI-AOI high-speed inspection…
Read more
A Fresh-Produce Processor: Integrated Foreign-Object Removal and Quality Grading with Deep Learning
On a 6.5 t/h high-speed line, DaoAI AI-AOI separates true foreign objects from natural raw-material variation, lifting FO detection above 97% while holding false reject below 1%.
Read more
A Prepared-Vegetable Plant: Real-Time Edge Detection of Small Foreign Objects in Fresh-Cut Produce
For millimeter-scale targets like hair, insects and plastic fragments, a DaoAI lightweight model runs real-time inference at the line edge, holding detection above 98%.
Read more
A Food-Packaging Plant: Inline AI Full-Inspection of Seals, Labels and Date Codes
Replacing rule-based vision with AI-AOI, the plant inspects seal integrity, label placement and date codes in one pass at over 99.5% accuracy, cutting complaints by 61%.
Read more
A Restaurant Chain: SkyVision Open-Kitchen Behavior Supervision for Food Safety
SkyVision runs on existing cameras to recognize gloves, hair nets and violations, with real-time alerts and auditable records — no hardware swap.
Read more
A Produce Sorting Center: Ripeness Grading and Robotic Vision Picking
DaoAI robotic vision grades ripeness and performs flexible pick-and-place sorting, improving grading consistency, freeing manual labor and lowering bruising.
Read more
Unsupervised Anomaly Detection on Textured Surfaces: Learn Good Only, Localize Unseen Defects
A materials plant faced complex surface textures and an endless variety of defect shapes that defy sample collection. DaoAI APDT learns from good products only, holding image-level…
Read more
Low-contrast Coating Boundary Detection: Dominant-channel Enhancement, 95%+ Recall on Faint Edges
A coating web plant struggled with near-zero color difference between coating and substrate, leaving edges, skips and shrinkage nearly invisible in grayscale. DaoAI's dominant-channel…
Read more
Web Coating Pinhole/Agglomerate Detection: 3D Point-cloud Fusion, >50μm at Line Speed
At a coating web plant, pinholes and agglomerates are micron-scale 3D defects that 2D imaging cannot tell apart from stains. DaoAI's 3D point-cloud fusion detects >50μm defects…
Read more
Print Film/Label Web False-alarm Reduction: Learn Per-machine Normal Variation, −28% False Alarms
A print film plant runs many machines and materials, where normal color and pattern variation was flagged as defects by the old system. DaoAI learns the normal-variation baseline per…
Read more
Glass/Nonwoven Surface Defect Detection: Few-shot Go-live, 94%+ Recall
A materials plant had scarce, highly variable defect samples on glass and nonwoven. DaoAI combined few-shot learning with good-only anomaly detection for fast go-live, reaching 94%+…
Read more
Transfer Learning for Assembly Missing/Wrong-Part Defects: 28 Classes, 92%+ Detection
A consumer-electronics plant runs high-mix, low-volume lines where new products launch with almost no defect samples. With DaoAI transfer learning reusing prior models, new-product…
Read more
Body Logo & Silkscreen Inspection: APDT Goes Live on Just 1–20 Good Samples, 40%+ Faster
A home-appliance plant frequently changes body logos and silkscreen artwork, making defect samples hard to collect. DaoAI APDT positive-sample learning goes live on just 1–20 good…
Read more
High-SKU Label Print Inspection: 30-Minute Changeover, 99.5% Detection, <0.3% False Reject
A hardware-tools plant manages thousands of label SKUs with extremely frequent line changes. DaoAI builds a model from just 20–50 golden samples per SKU, completes changeover within 30…
Read more
Reflective Metal Housing Grading: Trained on Good Samples Only, 94% Fewer Escapes
A consumer-electronics plant struggles with imaging reflective metal housings and scarce defect samples. DaoAI goes live trained on good samples only, delivering zero-code changeover…
Read more
Injection-Molded Part Inspection: Flash, Sink and Color in One Pass, Fast Changeover
An injection-molding plant juggles many molds and color batches with messy appearance defects. DaoAI unifies detection of flash, sink marks and color variation via good-sample…
Read more
People's Daily Defines AI's Next Chapter: DaoAI's GM on Physical AI
People's Daily's frontier feature 'Physical AI is accelerating into the real world' defines AI's next evolution — from the virtual to the physical. DaoAI's GM Zhang Yu: real-world…
Read more
AOI false calls won't stop? It's the architecture, not the settings
From pixel matching to feature cognition — how DaoAI ends AOI false calls at the architecture level. Color-space matching inherently can't separate same-color parts; feature-space…
Read more
DaoAI joins Canada's CTA Hard Tech Cohort
DaoAI joins Canada's CTA Hard Tech Cohort and lands at Computex InnoVEX 2026 in Taipei this June, at the Canada Pavilion — showing what industrial AI looks like when the agent design…
Read more
AI AOI vs. rule-based AOI: don't overlook these hidden costs
An EMS cost comparison — the real hidden bill behind programming, false calls and downtime. AI AOI and rule-based AOI aren't solving the same problem the same way.
Read more
NVIDIA says this is AOI's future — DaoAI already built it
In late 2025 NVIDIA said traditional CNNs had hit a ceiling; a visual foundation model lifted PCB-defect accuracy from 93.84% to 98.51%. DaoAI has already built that approach into…
Read more
The "programming tax": why your SMT line is only 70% efficient
Pick-and-place runs at 40,000 CPH, yet the line idles 2–3 hours per NPI for AOI programming — the programming tax. Few-Shot Learning cuts setup from 118 minutes to 5, letting the line…
Read more
2026 video-surveillance forecast: the era of active response
The era of piling on hardware and passive monitoring is over. The keyword for 2026 video surveillance is active response — AI augmentation, edge compute, active deterrence — shifting…
Read more