Omron and Apprhythm Improve Development Efficiency with Decentralized Learning Technology 'DcX'
Omron Corporation and Apprhythm Co., Ltd. have successfully built a馬体 detection AI model adaptable to stable environments by applying the decentralized learning technology 'DcX,' developed by Omron Sinic X Corporation, to the equine behavior monitoring AI product 'aiba.' This method allows AI models to be upgraded in a secure environment without sharing on-site data, resolving challenges related to development time and costs.
📋 Article Processing Timeline
- 📰 Published: June 9, 2026 at 20:08
- 🔍 Collected: June 9, 2026 at 11:21
- 🤖 AI Analyzed: June 9, 2026 at 12:57 (1h 36m after Collected)
Omron Corporation and Apprhythm Co., Ltd. have successfully developed a method for creating equine detection AI models that maintain expected performance across diverse environmental conditions. This was achieved by leveraging 'Decentralized X (DcX),' a decentralized learning technology developed by Omron Sinic X Corporation, and applying it to Apprhythm's AI product, 'aiba.'
Generally, developing AI models for object detection and image recognition requires training on specific field data. This poses challenges such as concerns over data usage rights, information leakage, and increased development time and costs due to growing data volumes. DcX offers a new approach that allows AI enhancement in a secure environment without data sharing, by integrating only the AI models trained at each specific site. In this verification, DcX was applied to horse stables and racecourse settings, confirming the ability to generate operationally viable models rapidly despite environmental changes.
Key achievements include improved estimation of equine center-of-gravity. By incorporating insights from other locations, stable detection became possible even in low-light night conditions at stables.
Omron plans to utilize these findings as an AI model generation technology for its 'Agentic AI' domain, as defined in its 'SF 2nd Stage' roadmap.
Generally, developing AI models for object detection and image recognition requires training on specific field data. This poses challenges such as concerns over data usage rights, information leakage, and increased development time and costs due to growing data volumes. DcX offers a new approach that allows AI enhancement in a secure environment without data sharing, by integrating only the AI models trained at each specific site. In this verification, DcX was applied to horse stables and racecourse settings, confirming the ability to generate operationally viable models rapidly despite environmental changes.
Key achievements include improved estimation of equine center-of-gravity. By incorporating insights from other locations, stable detection became possible even in low-light night conditions at stables.
Omron plans to utilize these findings as an AI model generation technology for its 'Agentic AI' domain, as defined in its 'SF 2nd Stage' roadmap.
FAQ
DcX（非集中学習技術）とはどのような技術ですか？
DcXは、各現場で学習されたAIモデルのみを持ち寄り、その出力結果を教師として用いる蒸留技術を利用した非集中型の学習手法です。データを外部に共有せず、各現場のデータを秘匿したまま、異なる環境の特徴を単一のAIモデルに統合できます。
今回の検証でどのような課題が解決されましたか？
これまでデータ共有に伴う利用権利や情報漏洩への懸念からデータを持ち寄ることが困難でしたが、DcXの活用により、現場固有のデータを共有することなく馬体検出AIモデルの精度向上と開発負荷の低減を両立しました。
「aiba」とはどのようなサービスですか？
アプリズムが提供する、馬の行動や状態をAIで見守り、管理業務を支援するAIプロダクトです。環境変化により検出精度が低下する課題がありましたが、DcXの適用により安定した検出が可能になりました。
オムロンの今後の展開方針を教えてください。
中期ロードマップ「SF 2nd Stage」のAgentic AI領域に向け、追加学習に伴う開発・運用負荷を低減する技術としてDcXを展開し、全社的なAI技術基盤の強化と事業競争力の向上を目指します。
本検証での各社の役割分担はどうなっていますか？
アプリズムはAIモデル開発と環境提供、オムロンはプロジェクトマネジメント、オムロン サイニックエックスはDcXおよびAI技術の研究開発とアドバイザリーを担当しました。