eval_smolvla_sorting_action_expert_ft_121ep_step160000_B_OOD_v1
收藏资源简介：
该数据集是一个机器人控制数据集，使用LeRobot工具创建，专门用于机器人任务，特别是与排序相关的动作。数据集包含2个完整episodes，总计2676帧，对应1个任务。数据以parquet文件格式存储，并附带MP4格式的视频文件（总大小约200MB），帧率为30fps。数据集特征包括：动作空间（6维浮点数，控制肩部平移、肩部提升、肘部弯曲、手腕弯曲、手腕旋转和夹爪位置）、观测状态（6维浮点数，反映机器人关节位置）、两个图像观测（来自顶部摄像头和手腕摄像头的视频，分辨率480x640，3通道RGB），以及时间戳、帧索引、episode索引等元数据。该数据集适用于机器人学习、模仿学习或强化学习研究，旨在支持模型训练和评估，尤其是在排序任务的OOD（Out-of-Distribution）场景下。
This is a robotic control dataset developed with the LeRobot toolkit, specifically tailored for robotic tasks, particularly sorting-related manipulation actions. The dataset contains 2 complete episodes, totaling 2676 frames, corresponding to one single task. The data is stored in Parquet file format, accompanied by MP4-format video files with a total size of approximately 200MB and a frame rate of 30 fps. The dataset features include: an action space with 6-dimensional floating-point values that control shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation and gripper position; observation states (6-dimensional floating-point values reflecting the robot's joint positions); two image observations from the top-down camera and wrist camera, with a resolution of 480×640 and 3-channel RGB; as well as metadata such as timestamps, frame indices and episode indices. This dataset is suitable for robotic learning, imitation learning or reinforcement learning research, aiming to support model training and evaluation, especially in out-of-distribution (OOD) scenarios for sorting tasks.
数据集概述
- 数据集名称:
eval_smolvla_sorting_action_expert_ft_121ep_step160000_B_OOD_v1 - 许可协议: Apache-2.0
- 任务类别: 机器人学 (Robotics)
- 标签: LeRobot
数据集规模
| 指标 | 数值 |
|---|---|
| 总片段数 | 2 |
| 总帧数 | 2676 |
| 总任务数 | 1 |
| 数据文件大小 | 100 MB |
| 视频文件大小 | 200 MB |
| 帧率 (FPS) | 30 |
机器人类型
- 机器人: so_follower
数据集结构
- 数据文件格式: Parquet 文件，路径为
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频文件格式: MP4 文件，路径为
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4 - 代码库版本: v3.0
- 分割: 仅包含训练集（
train），覆盖第 0 到第 2 个片段
特征说明
| 特征名称 | 数据类型 | 维度 | 说明 |
|---|---|---|---|
action |
float32 | (6,) | 6 个关节动作：肩部旋转、肩部升降、肘部弯曲、腕部弯曲、腕部旋转、夹爪位置 |
observation.state |
float32 | (6,) | 机器人状态观测，与动作维度一致 |
observation.images.top |
视频 | (480, 640, 3) | 顶部摄像头图像，编码为 AV1，帧率 30 FPS |
observation.images.wrist |
视频 | (480, 640, 3) | 腕部摄像头图像，编码为 AV1，帧率 30 FPS |
timestamp |
float32 | (1,) | 时间戳 |
frame_index |
int64 | (1,) | 帧索引 |
episode_index |
int64 | (1,) | 片段索引 |
index |
int64 | (1,) | 索引 |
task_index |
int64 | (1,) | 任务索引 |
