IDEAS-Lab-Northwestern/sentinel-pnp-clutter-joint
收藏资源简介:
该数据集是一个机器人数据集,使用LeRobot工具创建,基于FrankaPanda机器人。数据集包含1070个episodes、1,699,175帧、12个任务和3,210个视频,数据以parquet格式分块存储(共2个块,每块1000个episodes)。视频帧率为30fps,图像特征包括左侧图像(image_left)、右侧图像(image_right)和腕部图像(wrist_image),均为256x256x3的视频格式,使用h264编码。状态特征(state)为8维浮点数,表示7个关节位置和夹爪位置;动作特征(actions)为8维浮点数,表示7个关节目标位置和夹爪命令。其他特征包括时间戳、帧索引、episode索引、索引和任务索引。数据集仅包含训练集,所有episodes用于训练。
This dataset is a robotics dataset created using the LeRobot tool, based on the FrankaPanda robot. It contains 1070 episodes, 1,699,175 frames, 12 tasks, and 3,210 videos, with data stored in parquet format in chunks (2 chunks total, each with 1000 episodes). The video frame rate is 30 fps, and image features include left image (image_left), right image (image_right), and wrist image (wrist_image), all in video format with dimensions 256x256x3 and h264 encoding. The state feature (state) is an 8-dimensional float representing 7 joint positions and gripper position; the action feature (actions) is an 8-dimensional float representing 7 joint target positions and gripper command. Other features include timestamp, frame index, episode index, index, and task index. The dataset only includes a training set, with all episodes used for training.




