aaawangge/eval_so101_bimanual_total_0

该数据集是一个用于机器人学的数据集，使用LeRobot框架创建。它包含178个训练episodes，总计117505帧数据，帧率为30fps。数据集采用Apache 2.0许可证。数据以parquet格式存储，总数据文件大小约为100MB，视频文件大小约为200MB。数据集专为双臂机器人（robot_type: bi_so_follower）设计，包含7个不同任务。数据集结构包括：动作数据（action），为12维浮点向量，控制左右机械臂的肩部、肘部、腕部和夹爪位置；观察状态（observation.state），同样为12维浮点向量，表示机器人关节位置；来自四个摄像头的图像观察（observation.images.left_left, left_top, right_front, right_right），每个视频帧分辨率为480x640，3通道彩色，使用AV1编解码器；以及补充信息如策略动作（policy_action）、干预标志（is_intervention）、状态（state）、收集策略ID（collector_policy_id）等。此外，还包含时间戳、帧索引、episode索引、任务索引等元数据。数据集仅包含训练集，所有episodes均用于训练。

This dataset is a robotics dataset created using the LeRobot framework. It contains 178 training episodes, totaling 117505 frames at 30 fps. The dataset is licensed under Apache 2.0. Data is stored in parquet format, with total data file size approximately 100MB and video file size approximately 200MB. The dataset is designed for a dual-arm robot (robot_type: bi_so_follower) and includes 7 distinct tasks. The dataset structure includes: action data (action), a 12-dimensional float32 vector controlling positions of left and right arm shoulders, elbows, wrists, and grippers; observation state (observation.state), also a 12-dimensional float32 vector representing robot joint positions; image observations from four cameras (observation.images.left_left, left_top, right_front, right_right), each video frame with 480x640 resolution, 3 color channels, using AV1 codec; and complementary information such as policy action (policy_action), intervention flag (is_intervention), state (state), collector policy ID (collector_policy_id), etc. Additionally, it includes metadata like timestamp, frame index, episode index, and task index. The dataset only contains a training split, with all episodes used for training.