oxford_iiit_pet

  • विवरण :

ऑक्सफ़ोर्ड-IIIT पेट डेटासेट एक 37 श्रेणी का पेट इमेज डेटासेट है जिसमें प्रत्येक वर्ग के लिए लगभग 200 छवियां हैं। छवियों में पैमाने, मुद्रा और प्रकाश व्यवस्था में बड़े अंतर हैं। सभी छवियों में नस्ल और प्रजातियों की जमीनी सच्चाई का एनोटेशन जुड़ा हुआ है। इसके अतिरिक्त, प्रशिक्षण विभाजन के लिए हेड बाउंडिंग बॉक्स प्रदान किए जाते हैं, जो सरल ऑब्जेक्ट डिटेक्शन कार्यों के लिए इस डेटासेट का उपयोग करने की अनुमति देता है। परीक्षण विभाजन में, बाउंडिंग बॉक्स खाली हैं।

विभाजित करना उदाहरण
'test' 3,669
'train' 3,680
  • फ़ीचर संरचना :
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),
})
  • फ़ीचर दस्तावेज़ीकरण :
विशेषता कक्षा आकार डीप्रकार विवरण
फीचर्सडिक्ट
फ़ाइल_नाम मूलपाठ डोरी
हेड_बॉक्स बीबॉक्सफ़ीचर (4,) फ्लोट32
छवि छवि (कोई नहीं, कोई नहीं, 3) uint8
लेबल क्लास लेबल int64
विभाजन_मास्क छवि (कोई नहीं, कोई नहीं, 1) uint8
प्रजातियाँ क्लास लेबल int64
  • पर्यवेक्षित कुंजियाँ ( as_supervised doc देखें): ('image', 'label')

  • चित्र ( tfds.show_examples ): समर्थित नहीं है।

  • उदाहरण ( tfds.as_dataframe ): गुम है।

  • उद्धरण :

@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",
}