open_images_v4

Open Images เป็นชุดข้อมูลของรูปภาพประมาณ 9 ล้านภาพที่ได้รับการใส่คำอธิบายประกอบด้วยป้ายกำกับระดับรูปภาพและกล่องขอบเขตของวัตถุ

ชุดการฝึกของ V4 ประกอบด้วย bounding box 14.6M สำหรับคลาสอ็อบเจกต์ 600 คลาสบนอิมเมจ 1.74M ทำให้เป็นชุดข้อมูลที่ใหญ่ที่สุดที่มีอยู่พร้อมคำอธิบายประกอบตำแหน่งอ็อบเจ็กต์ กล่องเหล่านี้ส่วนใหญ่วาดขึ้นเองโดยผู้ทำหมายเหตุประกอบมืออาชีพเพื่อให้มั่นใจถึงความถูกต้องและสอดคล้องกัน ภาพมีความหลากหลายมากและมักมีฉากที่ซับซ้อนซึ่งมีวัตถุหลายชิ้น (โดยเฉลี่ย 8.4 ภาพต่อภาพ) นอกจากนี้ ชุดข้อมูลยังได้รับการใส่คำอธิบายประกอบด้วยป้ายกำกับระดับภาพที่ครอบคลุมคลาสนับพัน

แยก ตัวอย่าง
'test' 125,436
'train' 1,743,042
'validation' 41,620
  • โครงสร้างคุณลักษณะ :
FeaturesDict({
    'bobjects': Sequence({
        'bbox': BBoxFeature(shape=(4,), dtype=float32),
        'is_depiction': int8,
        'is_group_of': int8,
        'is_inside': int8,
        'is_occluded': int8,
        'is_truncated': int8,
        'label': ClassLabel(shape=(), dtype=int64, num_classes=601),
        'source': ClassLabel(shape=(), dtype=int64, num_classes=6),
    }),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'image/filename': Text(shape=(), dtype=string),
    'objects': Sequence({
        'confidence': int32,
        'label': ClassLabel(shape=(), dtype=int64, num_classes=19995),
        'source': ClassLabel(shape=(), dtype=int64, num_classes=6),
    }),
    'objects_trainable': Sequence({
        'confidence': int32,
        'label': ClassLabel(shape=(), dtype=int64, num_classes=7186),
        'source': ClassLabel(shape=(), dtype=int64, num_classes=6),
    }),
})
  • เอกสารคุณสมบัติ :
ลักษณะเฉพาะ ระดับ รูปร่าง Dประเภท คำอธิบาย
คุณสมบัติDict
วัตถุ ลำดับ
บ็อบเจกต์/บีบ็อกซ์ คุณสมบัติ BBox (4,) ลอย32
วัตถุ/is_depiction เทนเซอร์ int8
วัตถุ/is_group_of เทนเซอร์ int8
วัตถุ/is_inside เทนเซอร์ int8
อ็อบเจกต์/is_occluded เทนเซอร์ int8
bobjects/is_truncated เทนเซอร์ int8
วัตถุ / ฉลาก ป้ายกำกับคลาส int64
วัตถุ / ที่มา ป้ายกำกับคลาส int64
ภาพ ภาพ (ไม่มี, ไม่มี, 3) uint8
รูปภาพ/ชื่อไฟล์ ข้อความ สตริง
วัตถุ ลำดับ
วัตถุ / ความมั่นใจ เทนเซอร์ int32
วัตถุ/ฉลาก ป้ายกำกับคลาส int64
วัตถุ/แหล่งที่มา ป้ายกำกับคลาส int64
object_trainable ลำดับ
object_trainable/confidence เทนเซอร์ int32
object_trainable/label ป้ายกำกับคลาส int64
object_trainable/source ป้ายกำกับคลาส int64
  • คีย์ภายใต้การดูแล (ดู as_supervised doc ): None

  • การอ้างอิง :

@article{OpenImages,
  author = {Alina Kuznetsova and
            Hassan Rom and
            Neil Alldrin and
            Jasper Uijlings and
            Ivan Krasin and
            Jordi Pont-Tuset and
            Shahab Kamali and
            Stefan Popov and
            Matteo Malloci and
            Tom Duerig and
            Vittorio Ferrari},
  title = {The Open Images Dataset V4: Unified image classification,
           object detection, and visual relationship detection at scale},
  year = {2018},
  journal = {arXiv:1811.00982}
}
@article{OpenImages2,
  author = {Krasin, Ivan and
            Duerig, Tom and
            Alldrin, Neil and
            Ferrari, Vittorio
            and Abu-El-Haija, Sami and
            Kuznetsova, Alina and
            Rom, Hassan and
            Uijlings, Jasper and
            Popov, Stefan and
            Kamali, Shahab and
            Malloci, Matteo and
            Pont-Tuset, Jordi and
            Veit, Andreas and
            Belongie, Serge and
            Gomes, Victor and
            Gupta, Abhinav and
            Sun, Chen and
            Chechik, Gal and
            Cai, David and
            Feng, Zheyun and
            Narayanan, Dhyanesh and
            Murphy, Kevin},
  title = {OpenImages: A public dataset for large-scale multi-label and
           multi-class image classification.},
  journal = {Dataset available from
             https://storage.googleapis.com/openimages/web/index.html},
  year={2017}
}

open_images_v4/original (การกำหนดค่าเริ่มต้น)

  • คำอธิบาย การกำหนดค่า : รูปภาพที่ความละเอียดและคุณภาพดั้งเดิม

  • ขนาดชุดข้อมูล : 562.42 GiB

  • รูป ( tfds.show_examples ):

การสร้างภาพ

open_images_v4/300k

  • คำอธิบาย การกำหนดค่า : รูปภาพมีขนาดประมาณ 300,000 พิกเซล ที่คุณภาพ 72 JPEG

  • ขนาดชุดข้อมูล : 81.92 GiB

  • รูป ( tfds.show_examples ):

การสร้างภาพ

open_images_v4/200k

  • คำอธิบาย การกำหนดค่า : รูปภาพมีขนาดประมาณ 200,000 พิกเซล ที่คุณภาพ 72 JPEG

  • ขนาดชุดข้อมูล : 60.70 GiB

  • รูป ( tfds.show_examples ):

การสร้างภาพ