Fourth IEEE Workshop on Coding for Machines

Multimedia signals such as images, video, audio, and 3D data have traditionally been compressed for human perception. However, with the rapid growth of edge AI, multimodal large models, autonomous systems, and next-generation wireless networks enabling large-scale machine-to-machine (M2M) communication, multimedia data is increasingly consumed by machines rather than humans.

This shift calls for a rethinking of conventional compression pipelines. Instead of optimizing solely for perceptual quality and bitrate, coding methods must account for downstream machine tasks such as detection, segmentation, tracking, recognition, reasoning, and multimodal understanding—while operating under constraints of bandwidth, latency, complexity, energy, privacy, and robustness.

This workshop brings together researchers from compression, computer vision, multimodal learning, and communication systems to explore algorithms, representations, systems, and standards for efficient multimedia coding optimized for machine intelligence, as well as joint human–machine use.

Topics of Interest

We welcome high-quality, unpublished contributions including (but not limited to):

• Theories and frameworks for coding for machines

• Task-aware rate–distortion–complexity optimization

• Joint compression and downstream task training

• Coding for multiple machine tasks and human–machine use

• Feature, latent, and token compression for vision and multimodal models

• Coding for Large Vision-Language and Multimodal Models

• Compressed-domain multimedia analysis and processing

• 3D coding for machine perception

• Real-time edge analytics and M2M communication systems

• Software–hardware co-design for efficient deployment

• Datasets and benchmarking for coding for machines

• Error resilience, robustness, privacy, and security in machine-oriented coding