디엠랩

Dmlab 데이터 세트에는 DeepMind Lab 환경에서 작동하는 에이전트가 관찰한 프레임이 포함되어 있으며 에이전트와 환경에 있는 다양한 개체 사이의 거리로 주석이 추가됩니다. 목표는 3D 환경에서 시각적 입력으로부터의 거리를 추론하는 시각적 모델의 능력을 평가하는 것입니다. Dmlab 데이터 세트는 6개 클래스의 360x480 컬러 이미지로 구성됩니다. 클래스는 각각 {close, far, very far} x {positive reward, negative reward}입니다.

나뉘다
'test' 22,735
'train' 65,550
'validation' 22,628
  • 기능 구조 :
FeaturesDict({
    'filename': Text(shape=(), dtype=string),
    'image': Image(shape=(360, 480, 3), dtype=uint8),
    'label': ClassLabel(shape=(), dtype=int64, num_classes=6),
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
파일 이름 텍스트
영상 영상 (360, 480, 3) uint8
상표 클래스 레이블 int64

심상

  • 인용 :
@article{zhai2019visual,
        title={The Visual Task Adaptation Benchmark},
        author={Xiaohua Zhai and Joan Puigcerver and Alexander Kolesnikov and
               Pierre Ruyssen and Carlos Riquelme and Mario Lucic and
               Josip Djolonga and Andre Susano Pinto and Maxim Neumann and
               Alexey Dosovitskiy and Lucas Beyer and Olivier Bachem and
               Michael Tschannen and Marcin Michalski and Olivier Bousquet and
               Sylvain Gelly and Neil Houlsby},
                              year={2019},
                              eprint={1910.04867},
                              archivePrefix={arXiv},
                              primaryClass={cs.CV},
                              url = {https://arxiv.org/abs/1910.04867}
                          }