arXiv:2512.09235v1 Announce Type: new 
Abstract: The split-inference paradigm divides an artificial intelligence (AI) model into two parts. This necessitates the transfer of intermediate feature data between the two halves. Here, effective compression of the feature data becomes vital. In this paper, we employ Z-score normalization to efficiently recover the compressed feature data at the decoder side. To examine the efficacy of our method, the proposed method is integrated into the latest Feature Coding for Machines (FCM) codec standard under development by the Moving Picture Experts Group (MPEG). Our method supersedes the existing scaling method used by the current standard under development. It both reduces the overhead bits and improves the end-task accuracy. To further reduce the overhead in certain circumstances, we also propose a simplified method. Experiments show that using our proposed method shows 17.09% reduction in bitrate on average across different tasks and up to 65.69% for object tracking without sacrificing the task accuracy.

تقدم دراسة جديدة طريقة فعالة لضغط الميزات لنماذج الذكاء الاصطناعي، باستخدام تطبيع Z-score لتحسين استرداد بيانات الميزات المضغوطة. يتم دمج هذه الطريقة في تطوير معيار ترميز الميزات للآلات (FCM) من قبل مجموعة خبراء الصور المتحركة (MPEG)، متجاوزة الطرق الحالية المستخدمة في المعيار قيد التطوير وتظهر تحسينات كبيرة في تقليل معدل البت ودقة المهام.

Un nuevo estudio presenta un método eficiente de compresión de características para modelos de inteligencia artificial, utilizando la normalización Z-score para mejorar la recuperación de datos de características comprimidas. Este método se integra en el desarrollo del estándar de codificación de características para máquinas (FCM) por parte del Grupo de Expertos en Imágenes en Movimiento (MPEG), superando los métodos de escalado existentes y demostrando mejoras significativas en la reducción de bitrate y la precisión de las tareas.

Une nouvelle étude présente une méthode de compression de caractéristiques efficace pour les modèles d'intelligence artificielle, utilisant la normalisation Z-score pour améliorer la récupération des données de caractéristiques compressées. Cette méthode est intégrée dans le développement de la norme de codage de caractéristiques pour machines (FCM) par le Groupe d'experts en images animées (MPEG), surpassant les méthodes de mise à l'échelle existantes et montrant des améliorations significatives en matière de réduction de débit et de précision des tâches.

A new study introduces an efficient feature compression method for artificial intelligence models, utilizing Z-score normalization to enhance the recovery of compressed feature data. This method is integrated into the developing Feature Coding for Machines (FCM) codec standard by the Moving Picture Experts Group (MPEG), surpassing the existing scaling methods and demonstrating significant improvements in bitrate reduction and task accuracy.

Efficient Feature Compression for Machines with Global Statistics Preservation

One More Thing in AI – Your Shortcut to AI Mastery

Efficient Feature Compression for Machines with Global Statistics Preservation

Was this article worth reading? Share it

One More Thing in AI

LucidQuery AI

CompressX

Bytefitz

Capte

VideoDigest

Ready to build your own newsroom?