fmb

  • Descrizione :

Il nostro set di dati è costituito da oggetti con aspetto e geometria diversi. Richiede abilità motorie multistadio e multimodali per assemblare con successo i picchetti su una tavola non fissa in una scena randomizzata. Abbiamo raccolto un totale di 22.550 traiettorie in due diversi compiti su un braccio Franka Panda. Registriamo le traiettorie da 2 viste globali e 2 viste dal polso. Ciascuna vista contiene sia la mappa RGB che quella di profondità.

Diviso Esempi
'train' 1.804
  • Struttura delle caratteristiche :
FeaturesDict({
    'episode_metadata': FeaturesDict({
        'episode_language_embedding': Tensor(shape=(512,), dtype=float32),
        'episode_language_instruction': string,
        'episode_task': string,
        'file_path': string,
    }),
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=float32),
        'discount': Scalar(shape=(), dtype=float32),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'language_embedding': Tensor(shape=(512,), dtype=float32),
        'language_instruction': string,
        'observation': FeaturesDict({
            'color_id': Scalar(shape=(), dtype=uint8),
            'eef_force': Tensor(shape=(3,), dtype=float32),
            'eef_pose': Tensor(shape=(7,), dtype=float32),
            'eef_torque': Tensor(shape=(3,), dtype=float32),
            'eef_vel': Tensor(shape=(6,), dtype=float32),
            'image_side_1': Image(shape=(256, 256, 3), dtype=uint8),
            'image_side_1_depth': Tensor(shape=(256, 256), dtype=float32),
            'image_side_2': Image(shape=(256, 256, 3), dtype=uint8),
            'image_side_2_depth': Tensor(shape=(256, 256), dtype=float32),
            'image_wrist_1': Image(shape=(256, 256, 3), dtype=uint8),
            'image_wrist_1_depth': Tensor(shape=(256, 256), dtype=float32),
            'image_wrist_2': Image(shape=(256, 256, 3), dtype=uint8),
            'image_wrist_2_depth': Tensor(shape=(256, 256), dtype=float32),
            'joint_pos': Tensor(shape=(7,), dtype=float32),
            'joint_vel': Tensor(shape=(7,), dtype=float32),
            'length': string,
            'object_id': Scalar(shape=(), dtype=uint8),
            'primitive': string,
            'shape_id': Scalar(shape=(), dtype=uint8),
            'size': string,
            'state_gripper_pose': Scalar(shape=(), dtype=float32),
        }),
        'reward': Scalar(shape=(), dtype=float32),
    }),
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
metadati_episodio CaratteristicheDict
metadati_episodio/lingua_episodio_embedding Tensore (512,) float32
metadati_episodio/istruzione_lingua_episodio Tensore corda
metadati_episodio/attività_episodio Tensore corda
metadati_episodio/percorso_file Tensore corda
passi Set di dati
passi/azione Tensore (7,) float32
passi/sconto Scalare float32
passi/è_primo Tensore bool
passi/è_ultimo Tensore bool
passi/è_terminale Tensore bool
passaggi/incorporamento_lingua Tensore (512,) float32
passi/lingua_istruzioni Tensore corda
passi/osservazione CaratteristicheDict
passi/osservazione/color_id Scalare uint8
passi/osservazione/eef_force Tensore (3,) float32
passi/osservazione/eef_pose Tensore (7,) float32
passi/osservazione/eef_torque Tensore (3,) float32
passi/osservazione/eef_vel Tensore (6,) float32
passaggi/osservazione/image_side_1 Immagine (256, 256, 3) uint8
passi/osservazione/immagine_lato_1_profondità Tensore (256, 256) float32
passaggi/osservazione/image_side_2 Immagine (256, 256, 3) uint8
passi/osservazione/immagine_lato_2_profondità Tensore (256, 256) float32
passi/osservazione/immagine_polso_1 Immagine (256, 256, 3) uint8
passi/osservazione/immagine_polso_1_profondità Tensore (256, 256) float32
passi/osservazione/image_wrist_2 Immagine (256, 256, 3) uint8
passi/osservazione/immagine_polso_2_profondità Tensore (256, 256) float32
passi/osservazione/joint_pos Tensore (7,) float32
passi/osservazione/joint_vel Tensore (7,) float32
passi/osservazione/lunghezza Tensore corda
passi/osservazione/id_oggetto Scalare uint8
passi/osservazione/primitivo Tensore corda
passi/osservazione/shape_id Scalare uint8
passi/osservazione/dimensione Tensore corda
passi/osservazione/state_gripper_pose Scalare float32
passi/ricompensa Scalare float32
  • Citazione :
https://doi.org/10.48550/arXiv.2401.08553