�鶹��Ʒ��Ƶ

Researchers from �鶹��Ʒ��Ƶ’s Multimedia Systems Lab have achieved a major milestone in the field of AI and multimedia processing. Ashan Perera, a Researcher at the Multimedia Systems Lab, and PhD Candidate Juan Merlos traveled to Dublin, Ireland, to present their work titled “FCM-RT: Real-Time Feature Coding for Machines” at the ACM Multimedia 2025 conference which is one of the world’s leading venues for multimedia research.

The demonstration marks the first real-time implementation of the ongoing MPEG Feature Coding for Machines (FCM) standard, a pioneering international standardization effort under the Moving Picture Experts Group (MPEG). FCM defines how features extracted from split machine learning models can be efficiently transmitted, enabling collaborative edge–cloud intelligence for computer vision tasks.

The �鶹��Ʒ��Ƶteam’s FCM-RT system brings this concept to life by achieving real-time performance in transmitting and decoding intermediate neural network features, bridging the gap between AI computation and communication systems. The demonstration showcases how feature-level compression can significantly reduce bandwidth demands while maintaining inference accuracy which is an essential step toward scalable and efficient AI deployment across devices.

The paper and demo, authored by Ashan Perera, Md Eimran Hossain Eimon, Juan Merlos, Velibor Adzic, Hari Kalva, and Borko Furht, are available through the ACM Digital Library at: