2024.08.28 - #6 - Sapiens, GaussianOcc, FAST-LIVO2, SOLiD-ALOAM, Berkeley Humanoid, NavGPT-2

[changh95]

Papers

• Sapiens: Foundation for Human Vision Models
  • 메타에서 나온 Human foundation model ㄷㄷㄷ
  • 2D pose estimation, body-part segmentation, depth prediction and normal prediction이 하나의 모델에서 가능

• GaussianOcc: Fully Self-supervised and Efficient 3D Occupancy Estimation with Gaussian Splatting
  • 코드 릴리즈: https://github.com/GANWANSHUI/GaussianOcc

nuscenes_demo.mp4

• DoubleTake: Geometry Guided Depth Estimation (Niantic)
  • https://github.com/nianticlabs/doubletake

360562546-aa2052df-79f4-43a8-ab24-d704660f228a.mp4

• AirSLAM: An Efficient and Illumination-Robust Point-Line Visual SLAM System

  • https://github.com/sair-lab/AirSLAM?tab=readme-ov-file
  • https://youtu.be/5OcR5KeO5nc?si=TmBCPF-4-nl5L_Py

• ACE: A Cross-platform Visual-Exoskeletons for Low-Cost Dexterous Teleoperation

  • https://ace-teleop.github.io

Teleop.Xarm.Stacking.6.mp4

Teleop.B1.4.mp4

Industry news

• Unitree G1 휴머노이드 로봇 양산 준비

  • https://youtu.be/FuNFr7V7KFQ?si=abMH85EhjU9BkGFQ

• Cursor AI

  • vscode + AI 기반 IDE
    • 굉장히 효율적이지만, 회사에서는 쓰지 마세요!
    • 코드베이스 전부가 OpenAI/Anthropic/Cursor로 넘어가고, 코드베이스 공유 옵션을 꺼도 프롬프트가 넘어갑니다.
  • https://stream.mux.com/r43MZq02SaHg7uih1xnZs5DPswDnIBFQtu1D6VISv4XE.m3u8?min_resolution=720p&redundant_streams=true

Spatial AI study

• 가짜연구소 '스페셜한 Spatial AI'
  • 이번 기수에는 Gaussian Belief Propagation, Coded SLAM
    • 다음 기수에는 Object / Semantic SLAM, Scene graph, NeRF/3DGS feature fusion
  • 신청 링크 (9/6까지!): https://lnkd.in/giT-dixU

Help needed!

• GLIM을 쓰고 계신다면, 저자분께서 오픈소스 활동을 계속 이어가실 수 있도록 도와주세요!
  • GLIM을 어떻게 쓰고 계신지 적어주세요
  • Posting the way your are using GLIM and related packages. koide3/glim#19

Jobs

• Davide Scaramuzza 교수님 랩실 PhD 3 포지션: Vision-based Robot Learning with Foundation Models, Low-Energy Vision with Event Cameras, Vision-Based Localization for Autonomous Inspection, Robust Control in Confined Spaces 링크

[U-AMC]

Papers

FAST-LIVO2: Fast, Direct LiDAR-Inertial-Visual Odometry

SOLiD-A-LOAM

BodyTransformer

[rubidiumchain]

(Pilippeshin)

Interesting Papers

Berkeley Humanoid: A Research Platform for Learning-based Control (UC Berkeley, July 2024)

Simple Survey of Actuator Vendor

ZeroErr

tmotor

Mintasca

Industry News

NVIDIA Project GR00T: Generalist Robot 00 Technology

Tenstorrent API

TT-NN is a Python & C++ Neural Network OP library.

[iminolee]

News

• CMU Vision-Language-Autonomy(VLA) Challenge
  
  : For the first year(2024), CMU will host the competition in simulation(Unity, Habitat). Real-Robot Challenge Starting in 2025

• NavGPT-2: Unleashing Navigational Reasoning Capability for Large Vision-Language Models
  
  : Code Release on 2024/08/27 : https://github.com/GengzeZhou/NavGPT-2

Archive

• FAR Planner: Fast, Attemptable Route Planner using Dynamic Visibility Update (https://github.com/MichaelFYang/far_planner)

위 시스템을 연구실 real robot에 올려서 테스트한 결과 (Full Video: https://youtu.be/S_YusbxDUd4)

• FAEL 도 테스트 예정... (https://github.com/SYSU-RoboticsLab/FAEL)

Real-world experiment video: video