yxgao/egodex-wds

--- task_categories: - robotics language: - en tags: - egocentric - robotics pretty_name: EgoDex size_categories: - 100K<n<1M --- ```python from datasets import load_dataset, Features, Value, Sequence, Video, Array2D, Array4D EGODEX_DOF = 69 features = Features({ "__key__": Value("string"), "__url__": Value("string"), "confidences.npy": Array2D(shape=(None, EGODEX_DOF), dtype="float32"), "intrinsic.npy": Array2D(shape=(3, 3), dtype="float32"), "json": { "episode_id": Value("int64"), "has_confidences": Value("bool"), "joint_names": Sequence(Value("string")), "llm_description": Value("string"), "llm_description2": Value("string"), "num_frames": Value("int64"), "num_joints": Value("int64"), "original_resolution": Sequence(Value("int64")), "target_resolution": Sequence(Value("int64")), "task": Value("string"), "which_llm_description": Value("int64"), }, "mp4": Video(), "transforms.npy": Array4D( shape=(None, EGODEX_DOF, 4, 4), dtype="float32" ), "txt": Value("string"), }) dataset = load_dataset("yxgao/egodex-wds", streaming=True, features=features) ```### 数据集元信息 task_categories: - 机器人学 language: - 英语 tags: - 自我中心视角（egocentric） - 机器人学 pretty_name: EgoDex size_categories: - 100K<n<100万 python from datasets import load_dataset, Features, Value, Sequence, Video, Array2D, Array4D # EgoDex数据集的自由度（Degrees of Freedom，DOF）固定为69 EGODEX_DOF = 69 # 定义数据集的特征结构 features = Features({ "__key__": Value("string"), "__url__": Value("string"), "confidences.npy": Array2D(shape=(None, EGODEX_DOF), dtype="float32"), "intrinsic.npy": Array2D(shape=(3, 3), dtype="float32"), "json": { "episode_id": Value("int64"), "has_confidences": Value("bool"), "joint_names": Sequence(Value("string")), "llm_description": Value("string"), "llm_description2": Value("string"), "num_frames": Value("int64"), "num_joints": Value("int64"), "original_resolution": Sequence(Value("int64")), "target_resolution": Sequence(Value("int64")), "task": Value("string"), "which_llm_description": Value("int64"), }, "mp4": Video(), "transforms.npy": Array4D( shape=(None, EGODEX_DOF, 4, 4), dtype="float32" ), "txt": Value("string"), }) # 以流式模式加载yxgao/egodex-wds数据集，并使用上述定义的特征结构 dataset = load_dataset("yxgao/egodex-wds", streaming=True, features=features)