KuphDev/remove_green_duck_pruned

该数据集是一个机器人学习数据集，使用LeRobot工具创建，专门用于机械臂控制或模仿学习任务。数据集包含53个完整的情节（episodes），总计47,271帧数据，帧率为30 fps。数据特征包括动作（action）和观测状态（observation.state），均表示为6维浮点数组，对应机械臂的关节位置（如肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置）。观测部分还包括顶部摄像头图像（observation.images.top），分辨率为720x1280，3通道彩色视频，编码为h264。此外，数据集包含时间戳、帧索引、情节索引、任务索引等元信息。数据集采用Apache 2.0许可证，机器人类型为so_follower，数据以parquet文件格式存储，并分割为训练集（涵盖所有53个情节）。该数据集适用于机器人策略学习、状态估计或视觉-动作映射等研究。

This dataset is a robotics learning dataset created using the LeRobot tool, specifically designed for robotic arm control or imitation learning tasks. It contains 53 complete episodes, totaling 47,271 frames with a frame rate of 30 fps. The data features include actions and observation states, both represented as 6-dimensional float arrays corresponding to joint positions of the robotic arm (e.g., shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions). The observation also includes top camera images (observation.images.top) with a resolution of 720x1280, 3-channel color video, encoded in h264. Additionally, the dataset includes metadata such as timestamps, frame indices, episode indices, and task indices. Licensed under Apache 2.0, the robot type is so_follower, and the data is stored in parquet file format, split into a training set (covering all 53 episodes). This dataset is suitable for research in robot policy learning, state estimation, or vision-action mapping.