To further evolve humanity

We are developing various latest artificial intelligence systems using deep learning latest technologies (VAE, CNN, R – CNN, LSTM) and neural networks.

Specifically, we are developing image recognition, video analysis, behavior analysis, gesture recognition, etc. We are also developing a consistent artificial intelligence system such as automatic creation of teacher data necessary for deep learning.
We are also developing object recognition technology using smart glasses and AR glasses, which can be applied to in-plant work, assembly simulation, training and rehabilitation, education, medical field and so on.

Movie contents

AI transformation system to propose the best clothes for those worn by the user

It is a demonstration of VirtualFashion which uses deep learning which can recognize things (goods) owned by users and transform them according to thing (goods).

By incorporating the AI element into the AR signage “Kinesys“, for example, it is possible to automatically recognize the toys which children have taken, convert it into an animated hero, clothes matched to the hat worn by customers who visited the clothing store AI signage “ARIA” capable of coordinating can be proposed.

Product inspection system that reduces artificial mistakes at the factory and improves quality

It is a demonstration of a system using deep learning to check whether products are screwed in the correct position when looking at products through Microsoft HoloLens.
Depending on the application, it can be applied not only to inspection work but also to maintenance inspection and manufacturing fields. In the future, I would like to be able to deal with various smart devices.

Detect scratches and dirt on products that are often overlooked. Product malfunction detection system

It is a system that automatically detects scratches and dirt on products using deep learning.
Regardless of scratches and dirt locations and sizes, it is possible to automatically detect small scratches and dirt on any product.

To reduce human error in monotonous work. A system that visualizes the routine work of a person

Using deep learning system to visualize routine work of people.
It is difficult to avoid human error on site where there are many monotonous work such as factory site and we think that it is a system that can be a powerful approach to fields where measures were strongly desired.

It is to prepare the teacher data that becomes neck by deep learning, but the necessary data here is only the movie of the fixed point camera. The time required for classification will be about 4 hours of 5000 sheets using our own tool. In addition, we are currently developing a system that parallelizes learning “multi-node system”, and we hope to further reduce learning time.

Use Deep Learning for Inventory Management / Logistics Management

Deep Learning

We will develop systems such as object recognition, behavior analysis, and motion analysis using deep learning.
Combined with smart glass, AR glass (HoloLens), etc., it is expected to be applied to new logistics management and stock management.

Teacher data creation method for deep learning (learning of objects held in hand)

It is scenery of teacher data creation through web camera by deep learning. When learning, you learn by turning the item in front of the WEB camera.
After learning, when you display the product on the WEB camera, you can determine what the product is.
Easy, but the application range is wide, it is a very powerful approach.

AI system development example

Currently we are developing various AI systems as below.

  • Recognition of individual objects (individual products) using deep learning
  • Development / sale of the world’s first AI signage “ARIA (Aria)” incorporating individual object recognition AI engine
  • Behavior recognition, behavior analysis of people using deep learning, planned development up to action prediction in the future
  • Development of inspection system using deep learning (Detection of dirt on products, detection of mis-mounting of screws and parts)
  • Development of VR (virtual reality) system / MR (mixed reality) system incorporating artificial intelligence
  • Development of a realistic pseudo experience 3D simulation system
  • Development of image recognition and video recognition system using deep learning
  • Development of Human Interface System using Deep Learning
  • Development of a system to automatically track and memorize motion patterns by capturing movement of people from camera images
  • Recognition and learning of motion patterns using neural network
  • Development of VR system incorporating machine learning (Oculus VR / HTC VIVE + Kinect V2)
    Automatically learn (machine learning) the movement of the user in the virtual space, automatically recognize the standard motion pattern, and move the individual user’s movement in the virtual space based on the motion pattern I will judge.
    If it is not standard movement, you can inform you by issuing an alarm in the virtual space. It can be expected to be applied in training, rehabilitation, work training and so on.