Core Technology

Service model

Assisting clients in using AI and integrating deep learning with the client’s domain know-how to collect, preprocess, convert, and analyze data, and constructing AI training and validation models. We provide complete AI solutions and help clients to introduce virtuous cycle of AI.

1. Data collection

Label massive amounts of data of all sorts for AI training


2. Data conversion

Use data cleaning, data integration, and data transformation to convert data into appropriate formats for data mining.

3. AI model construction

Construct a variety of AI models and model result verification

4.System integration

Software design, function development of AI application end and system integration

AI services

AOI flow chart

About AOI product applications

AI-AOI equipment for industrial functional textile setting

AIWin has developed a fabric AI-AOI technology for the textile industry of Taiwan. It can help user building AI models quickly by AIWin AI Algorithms and provides fast and accurate inspection quality to reduce human fatigue effectively.

  • AIWin Deep Learning and Computer Vision Technology
    Detection Rate: more than 95%

  • Max. Speed: 25m/min.
  • lAIWin Special Mechanism

  • lDetect Defect Type: more than 6 types

  • Optical Resolution: 0.14mm/pixels



  • Recognition technology uses AIWin’s exclusive deep learning technology with a recognition rate of 98.5% or above.

  • The miss rate is lower than 0.1% (manual miss rate 1-3%); the over-detection rate lower than 10%.

  • It has an inspection speed of up to 3600  balls /hour.

  • Exclusive surface defect detection technology.

  • Effective in detecting more than 28 kinds of defects.

  • It is also able to detect different kinds of bad logos.

  • Optical precision 0.06mm/pixels

Optical image acquisition systems are installed along the shoelace production line. Cameras are used in combination with a deep learning AI image recognition system to detect defect on shoelaces’ surfaces.

Highlights of Technology

1. One-Shot Face Recognition technology – Recognition with only one photo file

2.Face recognition rate of up to 99.80% in the public database Labeled Faces in the Wild (LFW), Megaface over 98%.

3. The feature matching method does not need to save face files, solving worries about privacy intrusions.

4. Feature matching just takes 0.03 seconds to compare a person’s face. It takes only 0.1 second to compare an entire 720p picture.

5. The API can be called across different operating systems

6. Uses Nvidia GPU acceleration technology

IVS flow chart

About IVS product applications

Using license plate recognition, the whole parking lot can be managed with smart and automated technology. This can not only reduce operations cost but also car owners’ troubles (no manual collection of magnetic fobs). It also features software integration to accomplish data management, user blacklists, etc.

Human pose estimation is the analysis and solving of problems in behaviors and poses by locating the skeleton position with deep learning. It includes crowd monitoring, abnormal behavior monitoring, sleeping pose monitoring, fall monitoring, athlete movement analysis, etc.

The core function of facial recognition technology is to determine identification through images. AI and facial recognition algorithms are used to detect facial features and to analyze gender, age group, emotions. Related applications include blacklist detection, VIP recognition, door access control, unmanned shops, customer group analysis to retail businesses, etc.

AI image recognition technology is used to solve industrial safety issues, reducing the costs of related personnel. Applications include door access control, workers safety equipment inspection, site monitoring and abnormal conditions alarm, dangerous behaviors alarm, etc.