Test is automatically performed

using images prepared in Step 1.

You don’t need to adjust parameters

for differential inspections.

When a good product is judged

as defective, AI gives each image

a correlation score to visualize

the degree of overdetection. This

facilitates selecting images that need

to be learned to reduce overdetection.

5

Prepare images

1

2

3

Create model

Check results

Good

product

image

Defective

product

image

AI makes it easy to avoid overdetection

Three quick steps on the settings screen guide the user through the process of creating the good product model with

the minimum number of images.

Image A:10

Image B:150

Correlation score

AI model

AI Fine Matching identifies a future

that is not included in good products

as a defect.

AI learns images of good products

with variations, and generates an AI

model.

Every time an inspection is carried

out, AI reconstructs a model that is

presumed to be a good product.

AI extracts a difference between the

reconstructed good product image

and a captured image to identify a

defect, reducing overdetection.

Although standard AI processing

requires a huge number of

images for learning, the FH

Series requires only 100 to 200

images.

The system suggests images to

learn, helping to complete the

good product model.

*1. “Patent pending” means that we applied for a patent in Japan, and “Patented” means that we obtained a patent in Japan. (as of May 2022)

Shifted to upper right

Shifted to lower left

Reconstructs a good

product image considering

different views of holes

Captured image

AI model

Reconstructed

good product image

*1

PATENT PENDING

AI makes it easy

AI makes it easy

AI reduces overdetection

AI Fine Matching