HACARUS Inspect for Visual InspectionSPECTRO

AI Powered Inspection – Enabling Superior Results, even with limited training data.

Based on our proprietary AI technology, SPECTRO delivers faster and more accurate results, even with small amounts of training data.
Where traditional AVI / AOI systems are prone to false positives, SPECTRO excels and enables factory automation by vastly reducing the amount of reclassification needed by human inspectors.

SPECTRO Product Line-up

Use Case Examples

Defect Detection

Defect Detection

Detecting defects in a wide range of materials, including wood, metal, and fiber, using only normal product specifications.

Examples: Detection of flaws, dents, defects, and breakage

Missing parts & Misalignment Detection

Missing parts & Misalignment Detection

Detecting defects related to missing components or misassembled assemblies during inspection.

Examples: Detection of missing parts, misalignment of seals, and other part defects

Pattern Matching

Pattern Matching

Detecting defects related to deviations from the specified object pattern in addition to inspecting object uniformity.

Examples: Soldering inspection & identification of circuit board breakage points

Examples of Inspection Subjects

Printing condition

Printing condition

Printer cartridges

Printer cartridges

Metal plating

Metal plating

Counting number of parts

Counting number of parts

PCB

PCB

Solar panels

Solar panels

Wood parts

Wood parts

Inspection Performance

Visual Inspection for Solar Cells

Listed below is a comparison between our Visual Inspection AI, a classifier (SVM), and a Deep Learning method (CNN) for defect detection in solar cells. Compared to the other methods, our AI performed with higher accuracy and speed, even when the training data was limited.

* Sergiu Deitsch et al. : Automatic Classification of Defective Photovoltaic Module Cells in Electroluminescence Images. 2019
SVM CNN SPECTRO
training data size 800 models 800 models 60 models
Training time 30 min. 5 hr. 19 sec.
Inference time 8 min 20 sec. 10 sec.
Accuracy 85% 86% 90%

Subscribe to our newsletter

Click here to sign up