도시 경관

  • 설명 :

Cityscapes는 시맨틱 분할, TODO(인스턴스 레벨 분할) 및 스테레오 쌍 불일치 추론을 비롯한 여러 비전 작업에 대한 지상 진실뿐만 아니라 연중 다양한 시기에 50개 도시의 다양한 도시 거리 장면으로 구성된 데이터 세트입니다.

분할 작업(기본 분할, 'cityscapes/semantic_segmentation'을 통해 액세스 가능)의 경우 Cityscapes는 트레이닝(2975), 검증(500) 및 테스트(1525) 세트로 사전 분할된 1024 * 2048 해상도의 5000개 이미지에 대해 조밀한 픽셀 수준 주석을 제공합니다. 세그먼테이션 작업을 위한 레이블 주석은 주행 장면 인식 중에 일반적으로 발생하는 30개 이상의 클래스에 걸쳐 있습니다. 자세한 라벨 정보는 https://github.com/mcordts/cityscapesScripts/blob/master/cityscapesscripts/helpers/labels.py#L52-L99 에서 확인할 수 있습니다.

Cityscapes는 또한 'train_extra' 분할에서 19998 이미지에 대한 거친 입자 세분화 주석('cityscapes/semantic_segmentation_extra'를 통해 액세스 가능)을 제공하며 사전 학습/데이터가 많은 모델에 유용할 수 있습니다.

세분화 외에도 도시 경관은 일반 분할과 추가 분할(각각 'cityscapes/stereo_disparity' 및 'cityscapes/stereo_disparity_extra'를 통해 액세스 가능) 모두에서 시차 추론 작업을 위한 스테레오 이미지 쌍과 실측 정보를 제공합니다.

무시한 예:

  • 'cityscapes/stereo_disparity_extra'의 경우:
    • troisdorf_000000 000073 {*} 이미지(차이 맵 없음)
  • 추가 문서 : 코드가 있는 논문에서 탐색

  • 홈페이지 : https://www.cityscapes-dataset.com

  • 소스 코드 : tfds.datasets.cityscapes.Builder

  • 버전 :

    • 1.0.0 (기본값): 릴리스 정보가 없습니다.
  • 다운로드 크기 : Unknown size

  • 수동 다운로드 지침 : 이 데이터 세트는 원본 데이터를 download_config.manual_dir에 수동으로 download_config.manual_dir 해야 합니다(기본값은 ~/tensorflow_datasets/downloads/manual/ ).
    https://www.cityscapes-dataset.com/login/ 에서 파일을 다운로드해야 합니다(이 데이터 세트는 등록이 필요함). 기본 구성(semantic_segmentation)의 경우 'leftImg8bit_trainvaltest.zip' 및 'gtFine_trainvaltest.zip'을 다운로드해야 합니다. 다른 구성에는 추가 파일이 필요합니다. 자세한 내용은 코드를 참조하세요.

  • 자동 캐시 ( 문서 ): 아니요

  • 감독된 키 ( as_supervised 문서 참조): None

  • 그림 ( tfds.show_examples ): 지원되지 않습니다.

  • 인용 :

@inproceedings{Cordts2016Cityscapes,
  title={The Cityscapes Dataset for Semantic Urban Scene Understanding},
  author={Cordts, Marius and Omran, Mohamed and Ramos, Sebastian and Rehfeld, Timo and Enzweiler, Markus and Benenson, Rodrigo and Franke, Uwe and Roth, Stefan and Schiele, Bernt},
  booktitle={Proc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2016}
}

도시 경관/semantic_segmentation(기본 구성)

  • 구성 설명 : Cityscapes 시맨틱 분할 데이터 세트입니다.

  • 데이터세트 크기 : 10.86 GiB

  • 분할 :

나뉘다
'test' 1,525
'train' 2,975
'validation' 500
  • 기능 구조 :
FeaturesDict({
    'image_id': Text(shape=(), dtype=string),
    'image_left': Image(shape=(1024, 2048, 3), dtype=uint8),
    'segmentation_label': Image(shape=(1024, 2048, 1), dtype=uint8),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
image_id 텍스트
이미지_왼쪽 영상 (1024, 2048, 3) uint8
segmentation_label 영상 (1024, 2048, 1) uint8

도시 경관/semantic_segmentation_extra

  • 구성 설명 : train_extra 분할 및 거친 레이블이 있는 도시 경관 시맨틱 분할 데이터 세트.

  • 데이터세트 크기 : 51.92 GiB

  • 분할 :

나뉘다
'train' 2,975
'train_extra' 19,998
'validation' 500
  • 기능 구조 :
FeaturesDict({
    'image_id': Text(shape=(), dtype=string),
    'image_left': Image(shape=(1024, 2048, 3), dtype=uint8),
    'segmentation_label': Image(shape=(1024, 2048, 1), dtype=uint8),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
image_id 텍스트
이미지_왼쪽 영상 (1024, 2048, 3) uint8
segmentation_label 영상 (1024, 2048, 1) uint8

도시 경관/stereo_disparity

  • 구성 설명 : Cityscapes 스테레오 이미지 및 시차 지도 데이터세트입니다.

  • 데이터세트 크기 : 25.03 GiB

  • 분할 :

나뉘다
'test' 1,525
'train' 2,975
'validation' 500
  • 기능 구조 :
FeaturesDict({
    'disparity_map': Image(shape=(1024, 2048, 1), dtype=uint8),
    'image_id': Text(shape=(), dtype=string),
    'image_left': Image(shape=(1024, 2048, 3), dtype=uint8),
    'image_right': Image(shape=(1024, 2048, 3), dtype=uint8),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
disparity_map 영상 (1024, 2048, 1) uint8
image_id 텍스트
이미지_왼쪽 영상 (1024, 2048, 3) uint8
이미지 오른쪽 영상 (1024, 2048, 3) uint8

도시 경관/stereo_disparity_extra

  • 구성 설명 : train_extra 분할이 포함된 Cityscapes 스테레오 이미지 및 시차 지도 데이터세트입니다.

  • 데이터세트 크기 : 119.18 GiB

  • 분할 :

나뉘다
'train' 2,975
'train_extra' 19,997
'validation' 500
  • 기능 구조 :
FeaturesDict({
    'disparity_map': Image(shape=(1024, 2048, 1), dtype=uint8),
    'image_id': Text(shape=(), dtype=string),
    'image_left': Image(shape=(1024, 2048, 3), dtype=uint8),
    'image_right': Image(shape=(1024, 2048, 3), dtype=uint8),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
disparity_map 영상 (1024, 2048, 1) uint8
image_id 텍스트
이미지_왼쪽 영상 (1024, 2048, 3) uint8
이미지 오른쪽 영상 (1024, 2048, 3) uint8