2025.10.22 - #52 - SLAM 강의, WorldVLA, SceneDINO, VoT, InstantSfM, SAM3

[changh95]

SLAM 강의 출시!

• https://fastcampus.co.kr/data_online_slam
• 80시간어치 SLAM 강의 대체 어떻게 만들지.......

Interesting papers

WorldVLA: Towards Autoregressive Action World Model

• https://github.com/alibaba-damo-academy/WorldVLA
• The world model predicts future images by leveraging both action and image understanding, with the purpose of learning the underlying physics of the environment to improve action generation

SceneDINO: Feed-Forward SceneDINO for Unsupervised Semantic Scene Completion

• https://visinf.github.io/scenedino/
• Semantic Scene Completion (SSC) - 3D geometry + semantics 인퍼런스
• Multi-view self-supervision으로 학습

Visual Odometry with Transformers

• https://vladimiryugay.github.io/vot/

InstantSfM

• GPU 가속 BA를 통한 Structure-from-Motion
• COLMAP 보다 40배 빠름

SAM 3: Segment Anything with Concepts

• SAM 3: Segment Anything with Concepts
• Language prompt + instance segmentation + video tracking

[james-joobs]

책 리뷰 : AI 엔지니어링 (칩 후옌)

책 구매 : https://product.kyobobook.co.kr/detail/S000217939673

GitLink : https://github.com/chiphuyen/aie-book

목차