Observed arrival · 2026-11-13
EgoLens 拾迹: Searchable Memory for First-Person Video
A research prototype for selecting objects in first-person video, asking follow-up questions about them, and replaying the video and speech evidence behind answers.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Researchers and builders exploring egocentric video memory
- Worth noticing
- It distinguishes visible regions from confirmed physical identity and lets users correct SAM2 segmentation with added points.
Field notes
The method pairs point or box selection with a versioned SAM2 track, and says users can add points to correct segmentation errors. Its examples include clips of a tool, a utensil being moved, and two separately numbered objects; the source footage is identified as EgoLife data. The page also names the model components and states that its displayed Q&A records are completed runs rather than new model-generated answers.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue