2025.09.17 - #49 - ViPE, Maps for Autonomous Driving survey, TeraSim-World

[changh95]

Interesting papers

ViPE: Video Pose Engine for 3D Geometric Perception

• https://research.nvidia.com/labs/toronto-ai/vipe/
• ViPE efficiently estimates camera intrinsics, camera motion, and dense, near-metric depth maps from unconstrained raw videos.
• It is robust to diverse scenarios, including dynamic selfie videos, cinematic shots, or dashcams, and supports various camera models such as pinhole, wide-angle, and 360° panoramas
• runs at 3-5FPS on a single GPU for standard input resolutions
• designed to bridge the gap between classical and learning-based approaches. It combines the scalability and precision of a dense Bundle Adjustment (BA) framework, akin to SLAM, with the robustness of modern learned components.

[SunandBean]

Maps for Autonomous Driving: Full-process Survey and Frontiers

• 논문: https://arxiv.org/abs/2509.12632v1

Survey at A Glance

• 

illustrating the industry’s transition from high-definition maps to lightweight maps, and now to implicit, AI-driven model

• 

Typical pipeline of HD map production

• 

Comparison of Lite and HD map in terms of production pipeline and research focus

• 

A modern system for dynamic map maintenance through crowd-sourced vehicle data

• 

Implicit map-based methods for autonomous driving

• 

TeraSim-World: Worldwide Safety-Critical Data Synthesis for End-to-End Autonomous Driving

• 링크: https://wjiawei.com/terasim-world-web/
• 논문: https://arxiv.org/abs/2509.13164v1

TeraSim-World Pipeline

• 

Multi-View Video Generation Pipeline

• 

Case study at a roundabout in Ann Arbor, U.S. (42.31674°N, 83.70770°W

•