Paper reading

12.25

• RDT: https://github.com/thu-ml/RoboticsDiffusionTransformer
• 24.10.31 pi0: https://www.physicalintelligence.company/blog/pi0
• ALOHA + ACT: https://github.com/tonyzhaozh/act
• OpenVLA: https://github.com/openvla/openvla

• octo Open-Source Generalist Robot Policy: https://octo-models.github.io/

• 2023 具身智能 survey: https://eai-vc.github.io/

11.27

• Egomimic - RGB + handpose pretrain
  • https://egomimic.github.io/
• Dart: Vision Pro 纯虚拟环境收集数据
• 再看看 Pi0: 网络不适应夹爪，最后阶段 fine tune 一下
  • 夹爪 prompt?

11.20

• 清华的具身大模型: https://rdt-robotics.github.io/rdt-robotics/
  • https://arxiv.org/html/2410.07864?_immersive_translate_auto_translate=1
  • 视觉模型 Sig
• 跨具身模型 RTX 数据集: https://robotics-transformer-x.github.io/
  • 数据集下载: https://blog.csdn.net/qq_47572336/article/details/133592256?spm=1001.2014.3001.5502

11.18

• Diffusion 原理简述 (建议配合 GPT): https://www.mindspore.cn/tutorials/zh-CN/r2.4.0/generative/diffusion.html
• opencv-python 功能列表: https://opencv-python-tutorials.readthedocs.io/zh/latest/4.%20OpenCV%E4%B8%AD%E7%9A%84%E5%9B%BE%E5%83%8F%E5%A4%84%E7%90%86/4.5.%20%E5%BD%A2%E6%80%81%E5%8F%98%E6%8D%A2/

11.16

• Diffusion 生成 Floor Plan. https://arxiv.org/pdf/2410.11908v1
• https://github.com/luozn15/FloorplanGAN/blob/master/RPlan_EDA.ipynb
• Floor layout GAN: https://dl.acm.org/doi/abs/10.1145/3570991.3571057

11.13

• UMI pretrain 单一夹爪泛化 action: https://umi-gripper.github.io/
• fastumi: https://fastumi.com/
• 差不多的: https://robotutilitymodels.com/
• robot data scailing laws: https://data-scaling-laws.github.io/

11.6

• #[hand, rgb] RGB 重建手部 mesh: https://github.com/geopavlakos/hamer

10.29

• https://arxiv.org/abs/2404.08636: 探索大模型的 3D 泛化识别 Probing the 3D Awareness of Visual Foundation Models

10.26

• 形同我写的 IK 的 RelexedIK 2018: https://m.roboticsproceedings.org/rss14/p43.pdf
  • 优化用 scipy slsqp

10.19

• 无 teleoperation 的 mocap 灵巧数据采集: https://dex-cap.github.io/
  • 手部数据采集用的是电磁场手套