Computer Vision

101 summarised stories about Computer Vision, each linking back to the original source. Browse all topics →

Thursday, 9 July 2026

Meta Is Toying With the Idea of Smart Glasses That Record Everything, All the Time

TLDR 1 week ago

Meta is testing prototype smart glasses that continuously photograph the wearer's surroundings without requiring manual activation or displaying an LED indicator. The glasses would capture photos every few seconds and use the data to train AI systems, with internal disagreement at Meta over whether to store full images or only metadata on its servers. The shift represents a significant escalation from Meta's current smart glasses, which require voice or button activation to record and display a lit indicator when active.

Augmented Reality Computer Vision Data Privacy Meta

Incentivizing Temporal-Awareness in Egocentric Video Understanding Models

Apple ML Research 1 week ago

Researchers developed Temporal Global Policy Optimization (TGPO), a reinforcement learning algorithm that improves multimodal large language models' understanding of temporal sequences in egocentric videos by contrasting outputs from correctly ordered versus shuffled frames. TGPO was evaluated on five egocentric video benchmarks and consistently improved temporal grounding and causal coherence compared to prior RL-based approaches. The method enables MLLMs to better understand event ordering and narrative progression in first-person video understanding tasks.

Algorithm Innovation Computer Vision Multimodal AI Reinforcement Learning