Skip to content

2025.03.05 - #26 - SfM survey, Embodied AI simulator survey, PINGS, ESAM, Hier-SLAM++, S-Graphs 2.0, Fast3r #28

@changh95

Description

@changh95

Interesting papers

PINGS: Gaussian Splatting Meets Distance Fields within a Point-Based Implicit Neural Map

Image

EmbodiedSAM: Online Segment Any 3D Thing in Real Time (ESAM)

Hier-SLAM++: Neuro-Symbolic Semantic SLAM with a Hierarchically Categorical Gaussian Splatting

  • a novel and general hierarchical representation that encodes both semantic and geometric information in a compact form into 3D Gaussian Splatting, leveraging the capabilities of large language models (LLMs) as well as the 3D generative model
Image

S-Graphs 2.0 - A Hierarchical-Semantic Optimization and Loop Closure for SLAM

  • S-Graphs + semantic + floor-based hierarchical loop closure + floor/room-based hierarchical optimization
Image

Graph-Guided Scene Reconstruction from Images with 3D Gaussian Splatting

  • Select image pairs -> camera pose + matching -> octree initialization -> camera graph -> GS optimization
Image

LiDAR Registration with Visual Foundation Models

  • Wolfram Burgard, Davide Scaramuzza
  • DINOv2 features -> point descriptors
Image

Compression in 3D Gaussian Splatting: A Survey of Methods, Trends, and Future Directions

Image Image

Fast3R: 3D reconstruction of 1000+ images in a single forward pass

  • Fast3R achieves 251 FPS at its peak. 🔥 Try the demo with your images or video!
  • Fast3R ❤️LLM. Under the hood, Fast3R leverages:
    ⚡️ FlashAttention 2.0
    🚀 DeepSpeed ZeRO-2
    🔄 Positional Embedding Interpolation
    ⚙️ Tensor Parallelism

Image

Image

Metadata

Metadata

Labels

No labels
No labels

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions