The embedded deep learning framework
specialized for edge AI image recognitionKAIBER Lite
KAIBER Lite, featuring the image classification (CNN model) capability spun off from the standard edition of KAIBER, enables low-cost and easy embedded product development by specializing in edge AI development targeting on the ultrasmall comuter Raspberry Pi and a Windows application.
A touchless UI realizable with Raspberry Pi/Enabling DX
KAIBER Lite Features
01
Specialized in image classification (CNN)
Create a major model e.g. GoogleNet or your custom model rapidly
02
Raspberry Pi and Windows environments are supported
Easy-implementable as a dedicated inference engine (C library) is included as standard
03
The inference engine is gratuitously redistributable
Freely integratable to mass-production products as well as prototype products
04
It simplifies monocular camera based touchless UI development.
Touchless UI base module is included
The basic function including a pretrained model is attached as a library.It simplifies monocular camera based touchless UI development.
Demonstration videos
The Embedded Deep Learning FrameworkKAIBER
Deep learning specialized for embedded systems. Rapid and easy.
KAIBER is an embedded deep learning framework oriented toward "edge computing" which is being the key trend in the IoT field.
KAIBER features
01
Smallest-in-class (20K+) footprint is achieved.
Only a tiny module to embed
The inference executable module, fully separated from a training server, is easily integratable to various devices/applications.Smallest-in-class (20K+) footprint is achieved.
02
Commercial support tailored to your business is provided thanks to its purely in-house development.
Fully domestically-produced deep learning framework
To enable usability and flexible support, the entire code is developed from scratch.Commercial support tailored to your business is provided thanks to its purely in-house development.
03
Porting and customizing to novel platforms and devices can promptly be accommodated.
Portable to diverse architectures
The entire framework is developed by Java, and the inference executable module has a C language-implemented variant.Porting and customizing to novel platforms and devices can promptly be accommodated.
04
It allows you to try a wide variety of businesses, ranging from small-lot test-market to mass-production integration, with minimizing risks.
Please ask for customization.
Flexible billing model
The standard inference executable module is gratuitously distributable.It allows you to try a wide variety of businesses, ranging from small-lot test-market to mass-production integration, with minimizing risks.
Please ask for customization.
Deep Learning Edge EngineKAIBER engram
Realizing learning at the edge - Compact and flexible embedded machine learning solution
KAIBER engram features
01
Autonomous relearning
Autonomous relearning with being embedded in IoT GW / PLC / industrial vehicle
02
High-speed deep learning, plug-in
High-speed deep learning on Non GPU environments. Expandable plug-in for supporting novel GPUs and AI chips
03