2025 - Dataset

Aria Glass Gen-2/1



[OpenGaze]

The Visual Turing Test Redux


Dataset Labeling for 2020 - Aria Glass Hardware

ToolKit - On-device

2025 - Aria Gen 2

Tech report


Egocentric Vision, Intent Prediction, Anticipation, Multimodal Learning

HOT3D - 2025

Prototypes - only for Gen 1





Mesh Generations


đź“Ť 2025 - VertexRegen: Mesh Generation with Continuous Level of Detail

  • Controllable, ready-to-use mesh generation
  • Use a Coarse Mesh to estimate the global resolution initially, then gradually refine it to the local resolution


1996 - Microsoft Research - Progressive Meshes

  • Training data: Use edge collapse to compress the high-precision mesh into different levels

  • Generation process: Use a generative model to learn the inverse operation—vertex splitting

  • Thus, generation proceeds from coarse to fine, yielding a complete mesh at each step


2011 - High-quality passive facial performance capture using anchor frames



2025 - FantasyPortrait

  • đź“Ť Implicit facial expression Representations
  • DiT
  • Masked Cross-Attention



CUT3R


Navigation-level Scene Semantics


Sparse RGB / Depth / LiDAR (stream)
   ↓
Surface Fitting Module (Point cloud → implicit SDF)
   ↓
Continuous LOD generation
   ↓
4D Human Profile (geometry + temporal motion)
   ↓
Navigation / Control Integration
   - dynamic path planning
   - human-aware motion prediction




Topics


1. On-device Realtime Machine Perception (MP) Signals


Visual Inertial Odometry (VIO)

  • 6 Degrees of freedom (6DOF) within a spatial frame of reference using Visual Inertial Odometry (VIO)
  • This allows for seamless navigation and mapping of the environment



2. Eye Tracking


  • Including: gaze per eye, vergence point, blink detection, pupil center estimation, pupil diameter, corneal center, etc.
  • A deeper understanding of the wearer’s visual attention and intentions



3. Hand Tracking


  • In 3D space




References


2020 - LEMMA: A Multi-view Dataset for Learning Multi-agent Multi-task Activities



References