Skip to content
Move Beyond Tedious AOI Programming Start Inspecting.
Go from "One Board" to "Inspection Ready" with
AI Programming in 5 minutes

The DaoAI P Series is an AI-powered automated optical inspection (AOI) system designed for PCBA manufacturing. It reduces new product introduction (NPI) programming from hours to 5 minutes using a single reference board — no CAD files, component libraries, or specialized engineers required. The system uses Feature Cognitive Inspection to understand physical component concepts rather than matching pixel patterns, achieving 99%+ detection accuracy with up to 80% fewer false calls than traditional AOI. Fully on-premise with under 12ms inference latency, the P Series integrates into any SMT line via standard SMEMA conveyors.

3 Hours to 5 Minutes

Traditional AOI needs expert programmers.
DaoAI SMT Visual AI only needs one reference board.
1

Rapid Modeling

Industry: Open CAD, build a component library, align one by one.

DaoAI: Scan one reference board — learned automatically.

2

AI Component Mapping

Industry: Rely on CAD or hand-draw ROIs; rotated parts never align.

DaoAI: AI locates components, pins, and polarity in seconds.

3

AI-Generated Thresholds

Industry: Tune RGB rules one by one, iterate endlessly.

DaoAI: AI derives thresholds from the reference board.

Inspection Ready

Industry: Needs a programmer; false-call fatigue makes inspection a formality.

DaoAI: Operators ramp up fast; AI keeps learning from feedback.

Voice of the Customer · Real AOI Pain Points

First-hand feedback from operators of mainstream systems

PCBA Process Engineer

"Initial programming on a new board takes hours, and tuning takes even more. The whole workflow feels clunky.

Agent learns from a single reference board — new-board changeover in 5 minutes.

SMT Line Engineer

"The system mishandles rotated parts — change one and others rotate too, never matching the real footprint.

Multi-dimensional feature recognition handles any orientation — rotation and flips matched accurately.

Production Process Lead

"Solder-joint and bridge inspection drags throughput so badly we just turn them off.

Full-board solder-joint and ball inspection without impacting line cadence.

EMS SMT Manager

"High-mix orders are a nightmare — no time to micro-tune every detail.

No reprogramming for high-mix orders — fast changeover preserved.

Quality Engineer

"Operators see false alerts so often they just hit pass without looking. Inspection becomes a formality.

AI iterates with operator feedback; false-call rate shrinks over time.

AOI Operator

"Review only happens at an offline station after the board exits — no in-line, in-time comparison.

Live in-machine review and on-the-fly tuning.

Outpaces Mainstream AOI in Programming
01
Speed

Agent learns from one reference board — 5-minute changeover.

02
Handles Rotated Parts

Multi-dimensional features recognize any orientation — rotation, flips, accurately matched.

03
No CAD Library Needed

Agent learns from one reference board — no CAD, no library, no programming engineer.

Multi-Dimensional AI Detection · Overcoming Difficult Challenges

AI Component Alignment

Component surface contamination can degrade alignment accuracy, causing false positive. DaoAI uses AI algorithms for precise localization of contaminated components.

Same-Color Component Detection

When component body color matches PCB laminate, RGB color-based algorithms fail completely. DaoAI extracts multi-dimensional feature embeddings beyond RGB — reliably detecting components regardless of component color match.

DIP Solder Joint Classification (AI-Based)

DIP accept/reject criteria vary across customers, board types, and etcs. DaoAI trains a local AI classifier to your accept criteria — fast to deploy, high accuracy, no vendor dependency.

Black Inductor & Crystal Oscillator Mitigation

Chip inductors and crystal oscillators are notorious false-positive generators due to dense surface markings. DaoAI's multi-feature AI discriminator separates marking signatures from true defect signals — eliminating the root cause of false positives without rule-writing.

Detects All Defects:

Missing/Wrong Part Tombstone Polarity / Flipped Solder Bridge Misalignment OCR Foreign Object Insufficient Solder Solder Ball (Full-Board) Scratch (Full-Board)

Productivity Features

Program by part number & package type; sync across all instances AI-optimized imaging path — minimizes cycle time per board One-click panel alignment — no manual adjustment Automatic panel recognition: irregular, mirrored & combined panels Auto top/bottom side switching for double-sided inspection Auto-generated solder bridge detection regions Integrated bad-board rejection — automated, zero manual intervention

Traditional AOI VS. DaoAI AI AOI

Evaluation Criteria Traditional AOI Generic AI AOI DaoAI AI AOI
Training Methodology Manual rule-based programming General pre-trained model, not customized Trained specifically on your golden boards
Programming Time 3-5 hours per board Depends on AI engineer config Operator completes in 5 minutes
Data Security N/A Risk of IP exposure via cloud Fully on-premise — data never leaves factory
Optimization Fixed algorithms, no auto-iteration Improvements require vendor involvement Continuously learns from operator feedback
False Call Rate High false call rate; escape defects present Limited adaptability to specific processes Reduced by 80%; accuracy improves over time

Powered by DaoAI SMT Visual AI

The AI Brain Behind the Machine

Auto BOM Matching

Instantly sync scanned components to your BOM. No manual library creation.

Semantic False Call Filter

Intelligently distinguishes between 'Acceptable Variations' (oxidation, vendor font changes) and real defects.

Real-Time SPC Dashboard

Monitor defect trends, yield rates, and MES integration from a single interface.

Screenshot 2026-02-09 at 12.18.14
Screenshot 2026-02-09 at 12.18.42
Screenshot 2026-02-09 at 12.06.56

One Interface for Everything

  • Program new boards in same UI
  • Review flagged defects inline
  • Export reports to MES
  • Multi-language support (EN/CN)

Technical Features

Built on an industry-leading visual foundation model and trained on large-scale real-world SMT production data. Industrial-grade accuracy from day one.

Anomaly Scoring

AI-driven multi-dimensional feature extraction with anomaly scoring — not rule matching

Noise Immune

Model inference is feature-space, not pixel-space — immune to lighting drift and RGB noise

High Tolerance

High tolerance for optical variation, component lot-to-lot differences, and board warpage

1

Training Dataset: 1M+ Real-World Images

Factory-sourced image library spanning the full SMT component taxonomy. Continuously updated.

  • OK: conforming component reference images
  • NG: defect images across all failure modes
  • 4x4 multi-orientation augmentation
  • Active dataset expansion pipeline
2

AI Learns Directly from Real Samples

Employs Visual Foundation Model (VGG architecture) — convolutional neural network extracts features layer by layer for end-to-end deep learning.

  • Conv1: 224×224×64
  • Conv2: 112×112×128
  • Conv3: 56×56×256
  • Conv4: 28×28×512
  • Conv5: 14×14×512
  • fc6: 7×7×512
3

DaoAI AOI PCBA Visual AI

PCBA Visual AI Engine — Production-Grade Performance, Full-coverage defect taxonomy with <1ms per component inference speed.

  • Missing / Wrong / Misorientated component
  • Tombstoning / Physical damage / Polarity reversal
  • Surface scratches / Foreign material contamination

Engineered for Synergy.
Built for Flexibility.

Uncompromised Throughput
Technology

Scan technology icon

1 ms /component

Millisecond-level Inspection

Instantly detects missing parts, wrong components, and polarity on high-density boards without computing bottlenecks.

Parallel processing icon

45 cm²/s

Throughput

Perfectly matches the rhythm of High-Mix Volume SMT lines, eliminating post-reflow inspection bottlenecks.

Size comparison icon

CPU/GPU

Parallel Processing Architecture

GPU focuses on processing complex defects and OCR, while the CPU synchronously handles logic operations. True parallel processing ensures tasks do not wait for each other, maximizing hardware efficiency.

Elevate Your Yield with the P-Series

From compact offline stations to high-speed inline systems.

DaoAI P1/P2

DaoAI P1/P2

5MP Camera+ 15µm | 12MP Camera +10µm

Manual Load

DaoAI P3/P3D

DaoAI P3/P3D

12MP Camera | 10µm

Single Track/Dual Track

Equipment Specifications
Specification P1 / P2 (Offline) P3/P3D (Inline)
Ideal For High-Mix Volume High-Volume SMT Lines
Camera / Res 5MP (15µm) / 12MP (10µm) 20MP (10µm)
Imaging System Dual Telecentric Lens 20MP Telecentric Lens
Lighting Multispectral RGB 6-Channel High-intensity RGB Coaxial + multi-angle ring LED Ring
Max PCB Size 470 × 330 mm 500 × 325 mm
Conveyor Manual Load / Unload Single Track (P3) / Dual Track (P3D)
Ideal For
High-Mix Volume
Camera / Res
5MP (15µm) / 12MP (10µm)
Imaging System
Dual Telecentric Lens
Lighting
Multispectral RGB 6-Channel
Max PCB Size
470 × 330 mm
Conveyor
Manual Load / Unload
Ideal For
High-Volume SMT Lines
Camera / Res
20MP (10µm)
Imaging System
20MP Telecentric Lens
Lighting
High-intensity RGB Coaxial + multi-angle ring LED Ring
Max PCB Size
500 × 325 mm
Conveyor
Single Track (P3) / Dual Track (P3D)

Calculate Your ROI Savings

See how much you can save with DaoAI's AI-powered AOI solution

Cut Labor Costs

Save USD 28,000 – 42,000 per line every year.

Empower Operators

Move setup from 'Engineer-only' to 'Operator-ready' in 30 minutes.

Maximize Uptime

Turn hours of idle machine time back into profitable production.

Annual Savings
$28k $42k
Per line, every single year.

THE HIGH-MIX DILEMMA

Unraveling the complexities of modern manufacturing and the elusive promise of the Smart Factory.

Untitled.001-Feb-09-2026-07-54-27-8453-PM
Frequently Asked Questions

What is the 'NPI Paradox'?

It refers to the inefficiency where the time spent programming an AOI machine exceeds the actual production runtime of the batch. In High-Mix/Low-Volume (HMLV) manufacturing, this setup bottleneck often forces operators to skip inspection entirely for prototypes, increasing the risk of shipping defective boards.

How does DaoAI SMT Visual AI differ from Traditional AOI?

Unlike pixel-matching algorithms that rely on rigid component libraries and strict template matching, DaoAI utilizes deep learning generalization. Our system "sees" components like a human expert—recognizing parts based on features rather than exact pixel values—allowing it to handle slight variations in lighting, position, or vendor branding without triggering false alarms.

Do I need a 'Golden Board'?

Not a "perfect" pre-verified standard. You simply need one physical reference board from your current batch. The AI scans this unit to auto-learn component locations and visual characteristics instantly. This eliminates the need to source a flawless "Golden Sample" before you can even begin programming.

Can it work with CAD files?

Yes. The system seamlessly imports Pick-and-Place (XY coordinates) and BOM files. By combining CAD data with the visual scan, the AI automatically maps component locations and verifies polarity, reducing the programming time from hours to minutes.

Offline 2D AI AOI Machine

DaoAI P1 & P2

DaoAI P1& P2 leverages advanced visual AI to automate PCBA inspection programming. By learning directly from golden sample, it auto-detects components, generates bounding boxes, and defines defect thresholds — accelerating setup from hours to 5 minutes.

DaoAI product showcase

DaoAI P1/P2 Specifications

Designed for Speed, Accuracy, and Adaptability

DaoAI P1

5MP0 Camera /month

Resolution:15 µm

DaoAI P2

12MP0 Camera /month

Resolution:10 µm

DaoAI P1

DaoAI P2

Optical System

Camera
5 MP
12MP
Resolution
10 µm
15 µm
Lighting
Multispectral ultra-high-speed RGB six-channel lighting
Multispectral ultra-high-speed RGB six-channel lighting
Lens
Dual telecentric optical lenses
Dual telecentric optical lenses

Conveyor & PCB Size

PCB Size
0.5×50mm - 470×330mm
0.5×50mm - 470×330mm
PCB Thickness
0.5mm - 5mm
0.5mm - 5mm
Maximum PCB Warpage
Top25mm;Bottom110mm
Top25mm;Bottom110mm
Positioning Accuracy
±3mm
±3mm
Maximum Component Height
2.5mm
2.5mm

Inspection Capability

PCB Size
Missing components, Reversed components, Position offset, Polarity errors, Wrong components, Component damage, Bent leads, Foreign materials on PCB, Contaminated or solder-stained gold fingers, Glue overflow
Missing components, Reversed components, Position offset, Polarity errors, Wrong components, Component damage, Bent leads, Foreign materials on PCB, Contaminated or solder-stained gold fingers, Glue overflow
Solder Joint Inspection
Excess solder / insufficient solder, Tombstoning, Solder bridging, Cold solder / poor wetting, Lifted leads, Solder balls, Post–wave soldering defects (including through-hole components)
Excess solder / insufficient solder, Tombstoning, Solder bridging, Cold solder / poor wetting, Lifted leads, Solder balls, Post–wave soldering defects (including through-hole components)
Process Coverage
Supports inspection of solder paste, red-glue (SMT adhesive), and post–wave soldering PCB assembly quality
Supports inspection of solder paste, red-glue (SMT adhesive), and post–wave soldering PCB assembly quality

Additional Functions

Marking Function
PCB Fiducial Mark, Panel Mark, Break-off Tab Mark Recognition
PCB Fiducial Mark, Panel Mark, Break-off Tab Mark Recognition
Programming Function
CAD Data Import: Automated region generation and inspection parameter setupAutomatic Component Library Indexing
CAD Data Import: Automated region generation and inspection parameter setupAutomatic Component Library Indexing
Logging Function
Statistical Process Control (SPC) and Data Reporting
Statistical Process Control (SPC) and Data Reporting
Accessibility
Voice-Controlled Operation
Voice-Controlled Operation
 

Inspection and Auto-Configuration Capabilities

Body
Polarity IC Leads
Text

Capacitor

Resistor

Multilayer Chip Inductor

Aluminum Electrolytic Capacitor

Tantalum Capacitor

Resistor Array

Power Inductor

LED

SOT/SOD

SOIC/SOP/TSOP/TSSOP/MSOP

QFP

QFN

TO-263/TO-252