RayRoPE: Projective Ray Positional Encoding for Multi-view Attention

Original Tweet

---

[1/N] Rotary Position Embeddings (RoPE) are ubiquitous across transformers that process tokens from 1D, 2D, or 3D grids e.g. language, images, or videos. Our RayRoPE formulation extends these to multi-view transformers. Paper and code: https://rayrope.github.io/ https://x.com/shubhtuls/status/2019470333594468357/video/1

🔗 원본 링크

• https://rayrope.github.io/
• https://x.com/shubhtuls/status/2019470333594468357/video/1

미디어

---

🔗 Related

• 1n-rotary-position-embeddings-rope-are-ubiquitous-across-tra — 주제: Reconstruction, Transformer, Rope
• eag-pt-emission-aware-gaussians-and-path-tracing-for-indoor—281012 — 주제: Reconstruction, Rope
• iggt-instance-grounded-geometry-transformer — 주제: Reconstruction, Transformer
• introducing-shaper-a-method-for-robust-conditional-3d-shape-generation-from-casu — 주제: Reconstruction, Transformer
• vggt-slam-20-real-time-dense-feed-forward-scene-reconstructi — 주제: Reconstruction, Transformer

Tags

3D-Vision AI-ML