guloyy/SO101-eval2-bgr-expanded-v3-256

该数据集是一个用于机器人技术的数据集，专门针对so_follower机器人类型。它包含300个episodes、123,560帧和163个任务，数据以parquet文件格式存储，总数据文件大小为100 MB，视频文件大小为200 MB，帧率为30 fps。数据集的结构包括以下特征：动作（action）为6维浮点数组，表示机器人关节位置（如肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部滚动和夹爪位置）；状态观察（observation.state）同样为6维浮点数组，与动作对应；图像观察（observation.images.front）为256x256x3的视频数据，使用av1编解码器，无音频；此外，还包括时间戳、帧索引、集索引、索引和任务索引等元数据。数据集使用Apache 2.0许可证，并基于LeRobot工具创建，适用于机器人控制模型的训练和评估。

This dataset is designed for robotics, specifically for the so_follower robot type. It includes 300 episodes, 123,560 frames, and 163 tasks, stored in parquet format with a total data file size of 100 MB and video file size of 200 MB, at a frame rate of 30 fps. The dataset structure comprises features such as action (a 6-dimensional float32 array representing robot joint positions, including shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions); state observation (observation.state) as a 6-dimensional float32 array matching the action; image observation (observation.images.front) as video data with dimensions 256x256x3, using the av1 codec and no audio; and additional metadata including timestamp, frame index, episode index, index, and task index. Licensed under Apache 2.0, it was created using LeRobot and is suitable for training and evaluating robot control models.