시각화 : 알고 귀하의 데이터에 탐색
설명 :
dSprites는 6개의 실측 독립 잠재 요소에서 절차적으로 생성된 2D 모양의 데이터세트입니다. 이러한 요소는 컬러, 형상, 크기, 회전, X 및 스프라이트의 y 위치이다.
이러한 잠재성의 모든 가능한 조합은 정확히 한 번 존재하여 N = 737280 총 이미지를 생성합니다.
잠재 요인 값
- 색상: 흰색
- 모양: 정사각형, 타원, 하트
- 척도: [0.5, 1]에서 선형 간격으로 6개의 값
- 방향: [0, 2 pi]의 40개 값
- 위치 X: [0, 1]의 32개 값
- 위치 Y: [0, 1]의 32개 값
한 번에 하나의 잠재성을 변경하고(위치 Y에서 시작하여 위치 X 등) 고정된 순서로 이미지를 순차적으로 저장했습니다. 따라서 첫 번째 차원의 순서가 고정되어 해당 이미지에 해당하는 잠재 값으로 다시 매핑할 수 있습니다.
모든 픽셀 출력이 서로 다른지 확인하면서 가장 작은 단계 변경을 갖도록 의도적으로 잠재 값을 선택했습니다. 노이즈가 추가되지 않았습니다.
소스 코드 :
tfds.image.Dsprites
버전 :
-
2.0.0
(기본값) : 새로운 분할 API ( https://tensorflow.org/datasets/splits ) -
2.1.0
: 없음 릴리스 노트.
-
다운로드 크기 :
26.73 MiB
데이터 집합 크기 :
Unknown size
자동 캐시 ( 문서 ) : 알 수없는
분할 :
나뉘다 | 예 |
---|---|
'train' | 737,280 |
- 특징 :
FeaturesDict({
'image': Image(shape=(64, 64, 1), dtype=tf.uint8),
'label_orientation': ClassLabel(shape=(), dtype=tf.int64, num_classes=40),
'label_scale': ClassLabel(shape=(), dtype=tf.int64, num_classes=6),
'label_shape': ClassLabel(shape=(), dtype=tf.int64, num_classes=3),
'label_x_position': ClassLabel(shape=(), dtype=tf.int64, num_classes=32),
'label_y_position': ClassLabel(shape=(), dtype=tf.int64, num_classes=32),
'value_orientation': tf.float32,
'value_scale': tf.float32,
'value_shape': tf.float32,
'value_x_position': tf.float32,
'value_y_position': tf.float32,
})
감독 키 (참조
as_supervised
문서를 ) :None
그림 ( tfds.show_examples ) :
- 예 ( tfds.as_dataframe ) :
- 인용 :
@misc{dsprites17,
author = {Loic Matthey and Irina Higgins and Demis Hassabis and Alexander Lerchner},
title = {dSprites: Disentanglement testing Sprites dataset},
howpublished= {https://github.com/deepmind/dsprites-dataset/},
year = "2017",
}