haoanw/eval_pi0_20260426_182125

该数据集是一个机器人技术数据集，专门用于so_follower机器人类型。它包含11个完整的情节（episodes），总计9379帧数据，覆盖1个任务。数据以parquet文件格式组织，并包括以下特征：动作数据（包含6个浮点型关节位置，如肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置）、状态观测数据（与动作相同的6个关节位置）、图像观测数据（包括micro和macro两个视角的视频，每个视频分辨率为480x640，3通道，帧率为30fps，使用AV1编解码器），以及时间戳、帧索引、情节索引、索引和任务索引等元数据。数据集主要用于机器人控制、模仿学习或强化学习任务，支持从多模态观测中学习动作策略。

This dataset is a robotics dataset specifically designed for the so_follower robot type. It contains 11 complete episodes, totaling 9379 frames, and covers 1 task. The data is organized in parquet file format and includes the following features: action data (with 6 floating-point joint positions such as shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions), state observation data (same 6 joint positions as actions), image observation data (including videos from both micro and macro perspectives, each with a resolution of 480x640, 3 channels, 30fps frame rate, using AV1 codec), and metadata such as timestamp, frame index, episode index, index, and task index. The dataset is primarily used for robotics control, imitation learning, or reinforcement learning tasks, enabling learning of action policies from multimodal observations.