자동차196

Cars 데이터 세트에는 196개 자동차 등급의 16,185개 이미지가 포함되어 있습니다. 데이터는 8,144개의 훈련 이미지와 8,041개의 테스트 이미지로 분할되며 각 클래스는 대략 50:50으로 분할됩니다. 클래스는 일반적으로 제조사, 모델, 연도 수준입니다(예: 2012 Tesla Model S 또는 2012 BMW M3 쿠페).

나뉘다
'test' 8,041
'train' 8,144
  • 기능 구조 :
FeaturesDict({
    'bbox': BBoxFeature(shape=(4,), dtype=float32),
    'id': Text(shape=(), dtype=string),
    'image': Image(shape=(None, None, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=196),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
비박스 BBox기능 (4,) float32
ID 텍스트
영상 영상 (없음, 없음, 3) uint8
상표 클래스 레이블 int64

심상

  • 인용 :
@inproceedings{KrauseStarkDengFei-Fei_3DRR2013,
title = {3D Object Representations for Fine-Grained Categorization},
booktitle = {4th International IEEE Workshop on  3D Representation and Recognition (3dRR-13)},
year = {2013},
address = {Sydney, Australia},
author = {Jonathan Krause and Michael Stark and Jia Deng and Li Fei-Fei}
}