Visual Inspection AI
Quickly train and deploy AI models to automatically detect production defects—no technical expertise required.
-
Start quickly with a guided user interface on Google Cloud or our API
-
Train high-precision, domain-specific AI models to detect the tiniest defects
-
Minimal labeling effort due to active learning; start with only a few defect images
-
Run prediction models at the production line through Docker containers
Benefits
Run on-premises
Deploy high-performance inspection models at the network edge or on your factory floor.
Short time to value
Deliver significant ROI by reducing inspection costs, rework, and scrap and improving key quality metrics (e.g. escape rate, overkill rate, and yield).
Superior AI technology
Our top-ranked computer vision and machine learning technology lets you take on the most challenging inspection tasks.
Key features
Purpose-built, deep learning algorithm-based models for high-precision manufacturing inspection
Assembly inspection
Detects even the most subtle defects at various stages of the assembly process (wrong, misplaced, missing, rotated, or deformed components).
Cosmetic inspection
Locates even tiniest and most complex defects (dents, scratches, cracks, deformations, etc.) on any kind of surface.
Scale on-premises
Run models right on your shop floor with easily deployed Docker containers. Models are trained to meet your production quality requirements for escape and overkill rates.
Quick start capability
Start building models with only a few labeled images. Active learning will automatically suggest additional images for the operator to label and further improve the model’s performance.
"We cannot wait to roll the Assembly Inspection solution further across our extensive PCB manufacturing operations."
Sabcat Shih, Global Lead for Visual Inspection, FIH Mobile
Use cases
Explore common uses for Visual Inspection AI
Visual Inspection AI has been purpose-built for the production environment and addresses a wide range of use cases across the automotive, electronics, semiconductor, and industrial sectors.
Automotive manufacturers use Visual Inspection AI to inspect robot-welded seams for anomalies at the most critical structural joints of the chassis.
Electronics manufacturers use Visual Inspection AI to simultaneously inspect dozens of individual components on high volume printed circuit boards (PCBs) to detect missing, misplaced, or damaged components, screws, springs, and soldering issues.
Semiconductor manufacturers use Visual Inspection AI to detect and locate wafer defects, chip defects, or die cracks.
Compare features
Product comparison
AutoML
Visual Inspection AI
Typical use cases
-
General classification problems
-
Targeted for visual inspection tasks in manufacturing environments
On-premises deployment
-
Y
-
Y (through Docker container)
Manufacturing-specific models
-
N
-
Y (continuously improving due to domain-specific model optimization)
Imbalanced dataset handling
-
N (requires equal number of labeled defects and non-defects)
-
Y (can use a larger number of normal samples, plus a few labeled defects, which is typical in manufacturing)
Active learning
-
N
-
Y (identifies suspect examples quickly for human review and labeling)
Segmentation and localization
-
N
-
Y (pinpoint the area where the defect was found)
Missing component detection
-
N
-
Y
High resolution image support
-
N
-
Y (up to 100M pixels)
Automatic image alignment
-
N
-
Y (auto-align images from camera streams)
Image anomaly detection
-
N
-
Y
Typical use cases
-
General classification problems
On-premises deployment
-
Y
Manufacturing-specific models
-
N
Imbalanced dataset handling
-
N (requires equal number of labeled defects and non-defects)
Active learning
-
N
Segmentation and localization
-
N
Missing component detection
-
N
High resolution image support
-
N
Automatic image alignment
-
N
Image anomaly detection
-
N
Typical use cases
-
Targeted for visual inspection tasks in manufacturing environments
On-premises deployment
-
Y (through Docker container)
Manufacturing-specific models
-
Y (continuously improving due to domain-specific model optimization)
Imbalanced dataset handling
-
Y (can use a larger number of normal samples, plus a few labeled defects, which is typical in manufacturing)
Active learning
-
Y (identifies suspect examples quickly for human review and labeling)
Segmentation and localization
-
Y (pinpoint the area where the defect was found)
Missing component detection
-
Y
High resolution image support
-
Y (up to 100M pixels)
Automatic image alignment
-
Y (auto-align images from camera streams)
Image anomaly detection
-
Y
Pricing
Pricing
Please contact sales for pricing details
Partners
