Workshop - SEAS

Smart eyewear is emerging as one of the most challenging and impactful deployment frontiers for AI/ML. Unlike smartphones, glasses are worn continuously, sit millimeters from the human sensorium, and must operate within severe constraints on weight, thermal envelope, battery life, and end-to-end latency. At the same time, they integrate an unusually diverse sensor suite (RGB and event cameras, IMUs, microphone arrays, eye-tracking and EOG, depth sensors) and must deliver real-time, context-aware multimodal intelligence: vision-language understanding of the wearer's environment, speech enhancement and source separation for conversation in noise, gaze and attention modeling, gesture recognition, and AR content anchoring.

SEAS aims to be a dedicated venue for the AI-ML systems community working on smart eyewear and glasses-based AR. Topics of interest include:

• Efficient on-device foundation models for vision, audio, and multimodal tasks
• Aggressive quantization, pruning, and neural architecture search tailored to wearable constraints
• HW/SW co-design and novel accelerators for always-on sensing
• Three-tier architectures combining on-device, on-companion (phone/edge), and cloud inference
• Speech enhancement and hearing assistance on glasses
• Gaze-, IMU-, and context-aware adaptation
• Privacy-preserving and personalized on-device learning
• Benchmarks and datasets for glasses-mounted sensing

The workshop extends beyond the main conference tracks by focusing on the tightly coupled stack (from silicon to model to user experience) specific to eyewear form factors. It aims to connect a mature academic community in efficient and multimodal AI with a fast-growing industrial community building the next generation of intelligent glasses.