seriintan/eval_act_baseline_v4_round2

该数据集是一个机器人学数据集，由LeRobot工具创建。数据集包含12个训练集，总帧数为2236，总任务数为1，数据以parquet格式存储，总大小为100MB，视频文件大小为200MB，帧率为30fps。数据集使用so_follower类型的机器人，特征包括动作（action）、观测状态（observation.state）、来自前置和侧置摄像头的图像观测（observation.images.front和observation.images.side），以及时间戳、帧索引、集索引等元数据。动作和观测状态均为6维浮点数组，对应机器人关节位置（如肩部平移、肩部提升、肘部弯曲、手腕弯曲、手腕旋转和夹爪位置）。图像观测为视频格式，分辨率为480x640，3通道，使用av1编解码器。数据集适用于机器人控制和学习任务，采用Apache 2.0许可证。

This dataset is a robotics dataset created using the LeRobot tool. It contains 12 training episodes with a total of 2236 frames and 1 task. The data is stored in parquet format, with a total size of 100MB for data files and 200MB for video files, and a frame rate of 30fps. The dataset uses a so_follower robot type, and features include actions, observation states, image observations from front and side cameras (observation.images.front and observation.images.side), as well as metadata such as timestamps, frame indices, episode indices, etc. Actions and observation states are both 6-dimensional float arrays corresponding to robot joint positions (e.g., shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions). Image observations are in video format with a resolution of 480x640, 3 channels, using the av1 codec. The dataset is suitable for robotics control and learning tasks and is licensed under Apache 2.0.