licenta_dataset_wrist_finalpbn3

该数据集是一个机器人数据集，使用LeRobot工具创建，专为机器人控制和研究设计。数据集包含4个episodes，总计2586帧，帧率为30fps，数据以parquet格式存储，视频以mp4格式存储。数据集结构包括以下特征：动作（action）和观测状态（observation.state）均包含6个关节位置：肩部平移（shoulder_pan.pos）、肩部提升（shoulder_lift.pos）、肘部弯曲（elbow_flex.pos）、腕部弯曲（wrist_flex.pos）、腕部旋转（wrist_roll.pos）和夹爪位置（gripper.pos），数据类型为float32，形状为[6]。观测图像（observation.images.front）为前视摄像头视频，分辨率为480x640，3通道，视频编码为av1，像素格式为yuv420p，非深度图。此外，还包括时间戳（timestamp，float32）、帧索引（frame_index，int64）、episode索引（episode_index，int64）、索引（index，int64）和任务索引（task_index，int64）。数据集总数据文件大小为100MB，视频文件大小为200MB，分块大小为1000。数据集适用于机器人学习任务，如动作预测、状态观测和视觉控制。

This robotic dataset was constructed using the LeRobot toolkit, and is purpose-built for robotic control and research applications. The dataset includes 4 episodes, totaling 2586 frames with a frame rate of 30 fps. Tabular data is stored in Parquet format, while video content is saved in MP4 format. The dataset structure encompasses the following features: Both the action and observation.state fields contain 6 joint position values: shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, and gripper.pos, with a data type of float32 and a tensor shape of [6]. The observation image stream (observation.images.front) is captured by a front-facing camera, with a resolution of 480×640, 3 channels, using AV1 video encoding, YUV420p pixel format, and is not a depth map. Additional fields included in the dataset are timestamp (float32), frame_index (int64), episode_index (int64), index (int64), and task_index (int64). The total size of the tabular data files is 100 MB, the total size of the video files is 200 MB, and the chunk size is 1000. This dataset is suitable for robotic learning tasks such as action prediction, state observation, and visual control.