Fujitsu Kozuchi (code name)
Fujitsu AI Platform

Fast-track innovation – put the world’s latest cutting-edge technologies to the test

Japanese

This page introduces the research and development phase of Fujitsu Kozuchi. Product is available here.

Topics

  • February 22, 2024
    • Composite AI was added to AI Core Engines
  • February 15, 2024
    • Scene Change Detection was added to AI Core Engines
  • February 2, 2024
    • Quantum-Inspired Integrated Optimization was added to AI Core Engines
  • February 1, 2024
    • Large-Scale Point Cloud Data Integration was added to AI Core Engines
  • December 25, 2023
    • Fujitsu Auto Data Wrangling, Traffic Image Surveillance were added to AI Core Engines

What is the Fujitsu Kozuchi (code name) - Fujitsu AI Platform?

Fujitsu Kozuchi (code name) - Fujitsu AI Platform is a newly-developed concept designed to enable the fast testing of cutting-edge AI technologies developed by Fujitsu. The platform consists of two essential elements - AI innovation Components, which are packaged AI technology, and the AI Core Engines, Fujitsu's core AI technology. By using the platform, customers can achieve rapid business validation from a modest starting point.

AI Innovation Components

AI Innovation Components are packaged AI technologies designed to solve customer business challenges on a use case basis. In addition to cutting-edge AI technologies developed by Fujitsu, they provide standardized and coordinated AI technologies from OSS and partners, enabling the rapid realization of PoCs.

AI Core Engines

AI Core Engines are tools and software components that are based on Fujitsu's advanced AI technologies. Customers can also use AI core engines directly.

Fujitsu will deliver AI innovation Components and AI Core Engines one after another like "Uchide no kozuchi (magic mallet)" that you get whatever you wish for when you shake it. Through the Fujitsu Kozuchi (code name) - Fujitsu AI Platform, we aim to provide our customers with an important new pathway to kick-start PoCs, achieve growth and gain a competitive edge.

Customer Stories

FUJITSU I-Network Systems Limited
Visualizing variations of element work by worker or by work cycle by analyzing each factory worker's action and improving productivity by realizing work mastery earlier in comparison with standard operation times
Marukyu Co., Ltd.
Probe the impact of in-store digital signage by visualizing "Stop by," "Look," “Reach-out," and "Buy" behaviors

AI Innovation Components

Advanced AI features, models and tools that have been validated by pioneers in diverse industries such as manufacturing, retail and public sector are available.

Operation Process Analysis Optimizing production sites through operation process analysis focused on individual factory workers while reducing installation costs.
Product Defect Inspection Detecting defective products with high accuracy from image data and reducing the risk of bringing defective product to the market. Avoiding AI accuracy degradation due to changes in the manufacturing environment and improving product productivity and quality.
Defect Factor Analysis Improves product quality by analyzing infrequent defect data to determine the conditions of the manufacturing process under which defects occur and mitigating those factors.
Materials Discovery Enabling rapid discovery of new material candidates using accelerated quantum chemical simulations based on HPC and AI. The simulation results are then further enhanced using causal discovery AI. February 21, 2023 Press Release
Consumer Behavior Analysis Realizing more detailed visualization of purchasing behavior at real, supporting realization of smart stores that provide personalized customer experiences.
Cashier Fraud Monitoring Suppresses fraud due to accidental and intentional miscalculations in self-checkout areas through camera footage analysis and POS linkage to achieve loss prevention in short-staffed stores.
Auto-Advertisement Generation Using Generative AI to generate advertisement poster/video automatically with creative scripts, product key features, and poster/video design - reducing design costs with innovation and a high quality approach.
Suspicious Behavior Detection Early detection of suspicious behavior by using security cameras and surveillance systems. Realization of a safe and secure society.
3D Biomolecular Structure Prediction Conformational structural protein pathways are formed from cryo–electron microscopy images by applying our generative AI technology. Our goal is to improve the development process efficiency for medicines and vaccines. October 10, 2023 Press Release

AI Core Engines

The world's most accurate core engine enables quick and accurate output

Fujitsu AutoMLFujitsu's AutoML technology enables rapid generation of highly accurate machine learning models, including trial-and-error execution programs.VideoExperience Tool (Fujitsu Research Portal)
Fujitsu AI Ethics for FairnessAI learning data and fairness of decisions can be simply verified on a web browser and improved as needed.Experience Tool (Fujitsu Research Portal)
Causal DiscoveryQuick and comprehensive calculation of causal relationships under all conditions from vast amounts of data, discovery of individually causal relationships that are usually overlooked.May 17, 2023 Press Release
Wide LearningPresent reasons for AI judgements and field improving actions by exploring all the hypotheses from input table data.Experience Tool (Fujitsu Research Portal)
Conversational Generative AIVerify conversational generative AI, using your own business data.June 19, 2023 Press ReleaseMay 22, 2023 Press Release
ActlyzerEasily recognize complex human actions by using about 100 pre-learned basic action models and action recognition rules created by no-code UIRelated Site
TDA-Based Time Series Anomaly DetectionFujitsu’s TDA technology prevents and monitors serious incidents caused by failures and degradation, by analyzing the mechanism of time series changes.Related Site
Facial Expression RecognitionEnable communication support and health checks by capturing subtle and complex facial expressions from facial images, and so on.
Multi-Camera TrackingThis technology enables real-time tracking of the same person with multiple different cameras.
Synthetic Image GenerationEnabling users to generate high-quality synthetic image data with a desired style by multiple-condition controls.
Fujitsu Neuro-Symbolic ExplainerThis technology enables domain experts to validate image classifications made by AI, by using a neuro-symbolic approach to provide explanations for those classifications.
Camera Angle Change DetectionImmediately possible to detect when the camera direction is shifted! As well as the moment that stable operation of the image AI (such as preventing the operation from continuing while the camera is shifted) is possible.
Adversarial Example Attack DetectorFujitsu’s explainable learning technology detects adversarial example attacks, overcoming false identification of image recognition AI.
Fujitsu AI Ethics Impact AssessmentSimply enter the configuration AI system diagram into the tool and access a list of all ethical risks under European AI ethical guidelines.February 21, 2022 Press ReleaseRelated Site
Fujitsu Design Review AssistantWith our design document analysis technology powered by generative AI, users can easily detect and highlight errors in design documents such as omissions and erroneous statements.June 19, 2023 Press Release
Fujitsu Auto Data WranglingGame-changing advance to reduce data preparation in advance to AI application, via combination of data cleaning and enrichment.
Traffic Image SurveillanceProviding 27 functions of traffic image analysis functions in 5 categories, for efficient and low-cost safe traffic control and management.
Large-Scale Point Cloud Data IntegrationShift-changing approach removes the huge human effort of integrating point cloud data. Applicable to more than 1 million point cloud data.
Quantum-Inspired Integrated OptimizationSolve complex real-world combinatorial optimization problems by integrating unique quantum-inspired technology and advanced optimization algorithmsRelated Site
Scene Change DetectionSignificantly reduces man-hours required to check vacant space and inventory status by automatically detecting changes from surveillance camera images.
Composite AISolve customer problems that require combinatorial optimization technology through natural language interaction.October 11, 2023 Press Release

Powered by Fujitsu Kozuchi (code name) - Fujitsu AI Platform

Technologies developed in Fujitsu Kozuchi (code name) - Fujitsu AI Platform are used in a wide variety of different companies' services.

Deep Tensor is Fujitsu's proprietary deep learning technology that learns graph-structured data, which can represent various information relationships.

Endorsements

Endorsements from companies, universities, and organizations around the world.

Interested in testing the Fujitsu Kozuchi (code name) - Fujitsu AI Platform?

  1. Select the Innovation Components or Core Engines
  2. View demos
  3. Establish a trial environment
  4. Conduct a PoC

FAQ

    • A
      Fujitsu Kozuchi (code name) - Fujitsu AI Platform is a newly-redeveloped concept designed to enable the fast testing of cutting-edge AI technologies in various areas such as manufacturing, retail, and public sector.
    • A
      AI Innovation Components are packaged AI technologies designed to solve customer business challenges on a use case basis. In addition to cutting-edge AI technologies developed by Fujitsu, they provide standardized and coordinated AI technologies from OSS and partners, enabling the rapid realization of PoCs.
    • A
      AI Core Engines are tools and software components that are based on Fujitsu's advanced AI technologies.
    • A
      Fujitsu Kozuchi (code name) - Fujitsu AI Platform allows you to test Fujitsu Research's advanced AI technology quickly. By using the platform, customers can achieve rapid business validation from a modest starting point. Customization is also available by combining other providers' AI technology with Fujitsu's AI technology.
    • A
      Please contact us to discuss preparing a component and what knowledge is required for testing.
    • A
      There is no limitation on the number of users. Resource allocation based on the number of users involved will be arranged on an individual case basis.
    • A
      After the PoC environment set-up, access to the platform, data upload, and the use of components are all supported.
    • A
      New components and engines will be made available one after another.
    • A
      Ownership of the user data belongs to the users.
    • A
      The period from the initial application to the conclusion of the contract depends on how you conduct the test. We usually provide you with the test environment within a week after agreeing the contract. For customers outside of Japan, it may take longer to deal with export control procedures.

Connect with Fujitsu Research

LinkedIn
Twitter
YouTube
Top of Page