uiuc_d3field

  • وصف :

تنظيم مكتب المكتب والأواني وما إلى ذلك

ينقسم أمثلة
'train' 192
  • هيكل الميزة :
FeaturesDict({
    'episode_metadata': FeaturesDict({
        'file_path': Text(shape=(), dtype=string),
    }),
    'steps': Dataset({
        'action': Tensor(shape=(3,), 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': Text(shape=(), dtype=string),
        'observation': FeaturesDict({
            'depth_1': Image(shape=(360, 640, 1), dtype=uint16),
            'depth_2': Image(shape=(360, 640, 1), dtype=uint16),
            'depth_3': Image(shape=(360, 640, 1), dtype=uint16),
            'depth_4': Image(shape=(360, 640, 1), dtype=uint16),
            'image_1': Image(shape=(360, 640, 3), dtype=uint8),
            'image_2': Image(shape=(360, 640, 3), dtype=uint8),
            'image_3': Image(shape=(360, 640, 3), dtype=uint8),
            'image_4': Image(shape=(360, 640, 3), dtype=uint8),
            'state': Tensor(shape=(4, 4), dtype=float32),
        }),
        'reward': Scalar(shape=(), dtype=float32),
    }),
})
  • وثائق الميزة :
ميزة فصل شكل نوع D وصف
المميزاتDict
الحلقة_البيانات الوصفية المميزاتDict
Episode_metadata/file_path نص خيط المسار إلى ملف البيانات الأصلي.
خطوات مجموعة البيانات
الخطوات/الإجراء الموتر (3،) float32 إزاحة الروبوت من الإطار الأخير
الخطوات/الخصم العددية float32 الخصم إذا تم توفيره، الافتراضي هو 1.
الخطوات/is_first الموتر منطقي
الخطوات/is_last الموتر منطقي
الخطوات/is_terminal الموتر منطقي
الخطوات/language_embedding الموتر (512،) float32 تضمين لغة كونا. راجع https://tfhub.dev/google/universal-sentence-encoder-large/5
الخطوات/language_instruction نص خيط تعليم اللغة.
الخطوات/الملاحظة المميزاتDict
الخطوات/الملاحظة/العمق_1 صورة (360، 640، 1) uint16 كاميرا 1 لرصد العمق
الخطوات/الملاحظة/العمق_2 صورة (360، 640، 1) uint16 كاميرا 2 لرصد العمق.
الخطوات/الملاحظة/العمق_3 صورة (360، 640، 1) uint16 كاميرا 3 لرصد العمق.
الخطوات/الملاحظة/العمق_4 صورة (360، 640، 1) uint16 كاميرا 4 لرصد العمق.
الخطوات/الملاحظة/image_1 صورة (360، 640، 3) uint8 كاميرا 1 مراقبة RGB.
الخطوات/الملاحظة/image_2 صورة (360، 640، 3) uint8 كاميرا 2 RGB للمراقبة.
الخطوات/الملاحظة/image_3 صورة (360، 640، 3) uint8 كاميرا 3 مراقبة RGB.
الخطوات/الملاحظة/image_4 صورة (360، 640، 3) uint8 كاميرا 4 مراقبة RGB.
الخطوات/الملاحظة/الحالة الموتر (4، 4) float32 حالة المؤثر النهائي للروبوت
خطوات/مكافأة العددية float32 مكافأة إذا تم توفيرها، 1 في الخطوة النهائية للعروض التوضيحية.
@article{wang2023d3field,
  title={D^3Field: Dynamic 3D Descriptor Fields for Generalizable Robotic Manipulation},
  author={Wang, Yixuan and Li, Zhuoran and Zhang, Mingtong and Driggs-Campbell, Katherine and Wu, Jiajun and Fei-Fei, Li and Li, Yunzhu},
  journal={arXiv preprint arXiv:},
  year={2023},
}