oxford_iiit_pet

  • คำอธิบาย :

ชุดข้อมูลสัตว์เลี้ยง Oxford-IIIT เป็นชุดข้อมูลรูปภาพสัตว์เลี้ยง 37 หมวดหมู่ โดยมีรูปภาพประมาณ 200 ภาพสำหรับแต่ละคลาส รูปภาพมีขนาด ท่าทาง และแสงที่แตกต่างกันมาก ภาพทั้งหมดมีคำอธิบายประกอบตามความจริงที่เกี่ยวข้องของสายพันธุ์และสายพันธุ์ นอกจากนี้ ยังมีการจัดเตรียมกล่องขอบเขตส่วนหัวไว้สำหรับการแบ่งการฝึก ซึ่งช่วยให้สามารถใช้ชุดข้อมูลนี้สำหรับงานตรวจจับวัตถุแบบง่ายได้ ในการแบ่งการทดสอบ กล่องขอบเขตจะว่างเปล่า

แยก ตัวอย่าง
'test' 3,669
'train' 3,680
  • โครงสร้างคุณสมบัติ :
FeaturesDict({
    'file_name': Text(shape=(), dtype=string),
    'head_bbox': BBoxFeature(shape=(4,), dtype=float32),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=37),
    'segmentation_mask': Image(shape=(None, None, 1), dtype=uint8),
    'species': ClassLabel(shape=(), dtype=int64, num_classes=2),
})
  • เอกสารคุณสมบัติ :
คุณสมบัติ ระดับ รูปร่าง ประเภทD คำอธิบาย
คุณสมบัติDict
file_name ข้อความ เชือก
head_bbox คุณสมบัติ BBox (4,) ลอย32
ภาพ ภาพ (ไม่มี ไม่มี 3) uint8
ฉลาก ClassLabel int64
การแบ่งส่วน_mask ภาพ (ไม่มี ไม่มี 1) uint8
สายพันธุ์ ClassLabel int64
  • คีย์ภายใต้การดูแล (ดู as_supervised doc ): ('image', 'label')

  • รูปภาพ ( tfds.show_examples ): ไม่รองรับ

  • ตัวอย่าง ( tfds.as_dataframe ): หายไป

  • การอ้างอิง :

@InProceedings{parkhi12a,
  author       = "Parkhi, O. M. and Vedaldi, A. and Zisserman, A. and Jawahar, C.~V.",
  title        = "Cats and Dogs",
  booktitle    = "IEEE Conference on Computer Vision and Pattern Recognition",
  year         = "2012",
}