odp / README.md
KevinZonda's picture
Update README.md
82dbad1 verified
metadata
task_categories:
  - image-segmentation
language:
  - zh
  - en
  - ja
pretty_name: Optimised Danbooru Parsing
size_categories:
  - 10K<n<100K

Optimised Danbooru Parsing

An optimised version of DP (Danbooru Parsing). We merged serveral features to keep it simple, also, we convert it to numpy file so that can use np.load() load directly.

ODP is an anime face segmentation dataset (either the original DP). It has 11 labels:

ODP Label Id Description DP Label Id
0 Background 0
1 Face 1, 6
2 Eyebrows 2, 3
3 Eyes 4, 5
4 Ears 7, 8, 9
5 Nose 10
6 Mouth 11, 12, 13
7 Neck 14
8 Clothes 15, 16
9 Hair 17
10 Decoration 18, 15

Authors

Danbooru Parsing Authors, Xiang Shi & KigLand.