Skip to page content
SAVRN
Search Plan a project

Independent publisher

Lishan Yang

lishanyang

deep learning

Models in Library0
Datasets in Library1
Models on Hugging Face2
Followers—

Compare the published evidence

This profile connects the 1 model and dataset record shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.

Reported licenses: mit. Open a record to review its specific license and conditions.

Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.

Datasets

在 RoboTwin 2.0 里按官方轨迹做物理回放,导出每个 episode 前 132 帧 headcamera 的 ground-truth,供 OpenWAM 等表征的下游 probing 使用。 状态:已完成。 Agilex 2500 个 episode、Franka 2300 个 episode,共 4800 个 HDF5,全部通过结构检查,上传后都用 sha256 与 Hub 上的文件核对过。 episode 的范围就是 OpenWAM 仓库里 agilexrobotwin10smanifest.jsonl 和 frankarobotwin10smanifest.jsonl 列出的全部 episode,按 (task, episode) 对应。 episode.hdf5 里的数据都只来自 headcamera,没有 wrist、front 等其他相机,也没有视频和语言指令。手腕相机画面和其他补充数据在 episode extra.hdf5 里,见补充 GT。T 是该 episode 的帧数(不超过 132)。 - 分割:存的是调色板颜色,不是物体 ID,也没有颜色和物体名的对照表。颜色只在同一个 episode 里稳定,可以用来区分实例。上游 RoboTwin 的调色板只有 139 种颜色,遇到 label ID 更大的资产(例如 openlaptop)会报错,所以我们把调色板扩展到 256 种,前 139 种颜色保持原样。 - 几何检查:cam2worldgl 转成 OpenCV 约定后与 extrinsiccv 的差不超过 2e-7。把点云投影到同一帧的…

Publicly accessible mit