DPPE: Rethinking Camera-Based Positional Encoding for Scaling Multi-View Transformers
cs.CV, cs.AI
Submitted: 2026-06-30
Updated: 2026-11-03
Terminology
Sources
- LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias
- Cameras as Relative Positional Encoding
- MapAnything: Universal Feed-Forward Metric 3D Reconstruction
- Depth Anything 3: Recovering the Visual Space from Any Views
- Conditional Positional Encodings for Vision Transformers
- Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation
- Circle-RoPE: Cone-like Decoupled Rotary Positional Embedding for Large Vision-Language Models
- Rethinking Positional Encoding in Language Pre-training
- DeBERTa: Decoding-enhanced BERT with Disentangled Attention
- Spiral RoPE: Rotate Your Rotary Positional Embeddings in the 2D Plane
- Extending the Context of Pretrained LLMs by Dropping Their Positional Embeddings
- GTA: A Geometry-Aware Attention Mechanism for Multi-View Transformers
- RayRoPE: Projective Ray Positional Encoding for Multi-view Attention
- Unified Camera Positional Encoding for Controlled Video Generation
- URoPE: Universal Relative Position Embedding across Geometric Spaces
- ReDirector: Creating Any-Length Video Retakes with Rotary Camera Encoding
- CAT3D: Create Anything in 3D with Multi-View Diffusion Models
- MVImgNet2.0: A Larger-scale Dataset of Multi-view Images
- Stereo Magnification: Learning View Synthesis using Multiplane Images
- SpatialVID: A Large-Scale Video Dataset with Spatial Annotations
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs