eval_opentrajectorydit151_toycar
收藏资源简介:
该数据集是一个用于机器人模仿学习的评估数据集,由LeRobot创建。数据集包含10个片段(episodes),共5537帧,单个任务。机器人类型为so_follower,动作和观察状态均为6维(包括肩部旋转、肩部升降、肘部弯曲、腕部弯曲、腕部转动和夹爪位置)。数据包含侧视和手腕两个摄像头视角的640x480视频,帧率30fps,编码为AV1。数据集仅供训练使用,无验证/测试集划分。
This dataset is an evaluation dataset for robot imitation learning, created by LeRobot. It contains 10 episodes with a total of 5537 frames for a single task. The robot type is so_follower, with both action and observation states being 6-dimensional (including shoulder rotation, shoulder lift, elbow flexion, wrist flexion, wrist rotation, and gripper position). The data includes 640x480 videos from two camera perspectives (side view and wrist view), recorded at 30fps and encoded in AV1. The dataset is intended for training only and has no validation/test split.
- 数据集名称:edmos7/eval_opentrajectorydit151_toycar
- 许可证:Apache 2.0
- 任务类别:机器人技术(robotics)
- 标签:LeRobot
- 创建工具:LeRobot
- 数据集结构:
- 总集数:10集
- 总帧数:5537帧
- 任务总数:1个任务
- 数据分块:分块大小为1000帧
- 数据文件大小:100 MB(数据文件),200 MB(视频文件)
- 帧率:30 FPS
- 数据划分:训练集包含全部10集(索引0-9)
- 机器人类型:so_follower
- 特征:
- 动作(action):6维浮点数,包括肩部旋转、肩部升降、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置
- 观测状态(observation.state):6维浮点数,与动作特征相同
- 观测图像-侧方(observation.images.side):视频数据,分辨率480x640x3,AV1编码,30 FPS
- 观测图像-腕部(observation.images.wrist):视频数据,分辨率480x640x3,AV1编码,30 FPS
- 时间戳(timestamp):浮点数,单值
- 帧索引(frame_index):整数,单值
- 集索引(episode_index):整数,单值
- 全局索引(index):整数,单值
- 任务索引(task_index):整数,单值
- 数据路径:data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet(数据文件),videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4(视频文件)
- 引用信息:暂无提供BibTeX引用。



