- คำอธิบาย :
ชุดข้อมูลนี้มีตัวอย่างการเคลื่อนไหวของหุ่นยนต์ผลักดันประมาณ 44,000 ตัวอย่าง รวมถึงชุดการฝึกหนึ่งชุด (รถไฟ) และชุดทดสอบสองชุดของวัตถุที่เห็นก่อนหน้านี้ (ทดสอบ) และไม่เห็น (ทดสอบใหม่) นี่คือเวอร์ชัน 64x64 ขนาดเล็ก
รหัสที่มา :
tfds.datasets.bair_robot_pushing_small.Builderรุ่น :
-
2.0.0(ค่าเริ่มต้น): API แยกใหม่ ( https://tensorflow.org/datasets/splits )
-
ขนาดการดาวน์โหลด :
30.06 GiBขนาดชุดข้อมูล :
20.80 GiBแคชอัตโนมัติ ( เอกสารประกอบ ): ไม่
แยก :
| แยก | ตัวอย่าง |
|---|---|
'test' | 256 |
'train' | 43,264 |
- โครงสร้างคุณลักษณะ :
Sequence({
'action': Tensor(shape=(4,), dtype=float32),
'endeffector_pos': Tensor(shape=(3,), dtype=float32),
'image_aux1': Image(shape=(64, 64, 3), dtype=uint8),
'image_main': Image(shape=(64, 64, 3), dtype=uint8),
})
- เอกสารคุณสมบัติ :
| ลักษณะเฉพาะ | ระดับ | รูปร่าง | Dประเภท | คำอธิบาย |
|---|---|---|---|---|
| ลำดับ | ||||
| การกระทำ | เทนเซอร์ | (4,) | ลอย32 | |
| endefector_pos | เทนเซอร์ | (3,) | ลอย32 | |
| image_aux1 | ภาพ | (64, 64, 3) | uint8 | |
| image_main | ภาพ | (64, 64, 3) | uint8 |
คีย์ภายใต้การดูแล (ดู
as_superviseddoc ):Noneรูปภาพ ( tfds.show_examples ): ไม่รองรับ
ตัวอย่าง ( tfds.as_dataframe ):
- การอ้างอิง :
@misc{1710.05268,
Author = {Frederik Ebert and Chelsea Finn and Alex X. Lee and Sergey Levine},
Title = {Self-Supervised Visual Planning with Temporal Skip Connections},
Year = {2017},
Eprint = {arXiv:1710.05268},
}