Home   »   Product   »   SuaKIT


Product Description

SuaKit is a deep learning-based machine vision inspection software

that can be used in various manufacturing sites such as PCB, solar,

battery and steel.

SuaKit is a library based on actual image data generated from vari

ous industrial sites and is a proven software that's been verified by

world-famous companies such as Samsung and Hanhwa.

SuaKit, to this day, is undergoing learning and testing processes

based on various data sets in the industry and continues to improve


SuaKit offers three functions.

Suakit image02


Various defects that occur at manufacturing sites are detected through image analysis.

Inspection covers a range of product types that have varying surface shapes,

such as fiber, leather, solar panels and camera lenses.

The various products can be sorted by type.

Suakit image02


The software analyzes the image and classifies them according to the defined

features and by class. It can be used to film and sort the various objects at the

manufacturing sites.


Improved Inspection


Deep Learning-based inspection algorithms ensure higher accuracy than visual inspection.

Reduction of

Optimization Costs

After the initial learning, the optimization process will automatically proceed in the future.


Human Capital


As the solution enables automatic inspection with high accuracy, it can manage a number of

inspection equipments even with just a single employee. Customers can achieve efficient workforce

management by minimizing personnel.


We provide the best UX for the customer based on intuitive UI,

maximizing the usability and utility for the customer. From the labeling tool to the resulting output,

we have created a customer-friendly environment.

Utilizing GPU-specific processing languages such as CUDA and cuDNN,

we've maximized the processing speed of the deep learning algorithm.

We have developed a high-accuracy network based on know-how gained in various manufacturing fields,

such as textile, leather, LCD, PCB, solar energy and battery.


How many pieces of image data should I teach the machine?

It varies depending on the complexity of the image,

but starting with about 100 images per type of defect should be sufficient.

How long does it take to learn the network?

It depends on the number of images the machine is studying,

but it takes about 5 minutes to learn 1,000 sheets of 2048x2048 images.

How fast is the image data processed on the inspection line?

It depends on the image size and resolution,

but processing speed of up to 150m/min, based on dimensions of 2k*2k and 20㎛, is possible.




Contact us

If you have additional questions about the prices and/or products, please contact us by e-mail or phone.

Email : sales@sualab.com     /     Tel : 02-6264-0362