Features


Factory Automation
  • It supports “image fragment division classification”, and by training and recognizing the high-resolution product images by dividing each into small pieces, it is possible to identify the defective part of the product in detail.


  • It is very important to minimize the quantization error in a quality inspection system where the difference between good and bad products is small. Deep Runner minimizes quantization errors through floating-point operations in deep learning computation.


  • It supports high-quality uncompressed YUV images and does not go through the lossy compression process that degrades the image quality, so subtle differences in images can be identified.



Security System
Detect and eliminate harmful animals
Detect and eliminate harmful animals
High-resolution object recognition: Deep Runner web-based output
High-resolution object recognition: Deep Runner web-based output
  • In general, deep learning algorithms recognize small-sized images as small as 300x300 pixels, so it is difficult to recognize small objects in the image. Deep Runner supports high-resolution input so it can detect small objects in the image.


  • It provides various pre-trained object recognition functions. In this case, deep learning training is not required, and the desired object can be recognized with a few clicks and linked with the IT system.


  • Supports the function to automatically save images to disk when a specific object is found.


  • By supporting web-based screen output, real-time monitoring is possible remotely.


Car License Plate Recognition
Multi-model linkage recognition - license plate
Multi-model linkage recognition - license plate
  • By using the latest deep learning technology (recognition of multiple linked models), complex license plates can be recognized with the highest recognition rate.


  • If you can collect enough car license plate images, it is possible to make customized responses to various license plate structures and letter shapes for each country.


  • One small device suffices, and it does not require a separate compute providing high economy.
Smart Retail
Multi-Model Linkage Recognition - Gender/Age/Emotion
Multi-Model Linkage Recognition - Gender/Age/Emotion
Automatic alculation of smart vending machine products
Automatic alculation of smart vending machine products
  • By linking multiple learning models at a kiosk or smart sales stand, you can perform image recognition of various scenarios. For example, when a person appears, not only the position of the face, but also the age, gender, emotion, etc. can be automatically recognized sequentially.


  • It is easy to implement a system that automatically calculates (checkout) items to be purchased by pointing the camera at them. Anyone can easily learn to use it with images of the products being sold in advance.





Deep Runner does not require artificial intelligence personnel.
Anyone can use it easily by following the guide we provide.

Anyone can easily implement a complex scenario configuration by clicking on the web screen.

Even a general IT engineer can easily build a system that works with the existing IT system.

It is a small device and provides high cost-effectiveness by self-providing all AI functions.