---
dataset_info:
  features:
  - name: index_num
    dtype: int64
  - name: filename
    dtype: string
  - name: s3_uri
    dtype: string
  - name: pixiv__pid
    dtype: int32
  - name: clip__clip_aesthetic
    dtype: float32
  - name: pixiv__likeCount
    dtype: float64
  - name: pixiv__viewCount
    dtype: float64
  - name: pixiv__bookmarkCount
    dtype: float64
  - name: pixiv__aiType
    dtype: float64
  - name: pixiv__compound_score
    dtype: float32
  - name: caption
    struct:
    - name: alternate_tags_mid
      sequence: string
    - name: tags_back
      dtype: string
    - name: tags_front
      dtype: string
    - name: tags_front_fixed
      dtype: string
    - name: tags_mid
      dtype: string
  - name: aigccls__aigc_pred
    dtype: string
  - name: aigccls__is_likely-ai
    dtype: bool
  - name: dset__pixels
    dtype: float64
  - name: dset__from
    dtype: string
  - name: danbooru_id
    dtype: float64
  - name: danbooru__up_score
    dtype: float32
  - name: danbooru__down_score
    dtype: float32
  - name: danbooru__fav_count
    dtype: float32
  - name: danbooru__is_banned
    dtype: bool
  - name: danbooru__tag_string_artist
    dtype: string
  - name: danbooru__tag_string_character
    dtype: string
  - name: clip__twitter_aesthetic_v2
    dtype: float32
  - name: comps__comp_score_pred
    dtype: float32
  - name: resized__resized_s3_uri
    dtype: string
  - name: clip_aesthetic_2_5
    dtype: float32
  - name: danbooru__quality_tag
    dtype: string
  - name: twitter__favorite_count
    dtype: float64
  - name: twitter__hashtags
    dtype: string
  - name: twitter__compound_score
    dtype: float64
  - name: twitter__quality_tag
    dtype: string
  - name: dset__quality_tag
    dtype: string
  - name: dset__artist_occurence_count
    dtype: int64
  - name: dset__row_cap_dict
    struct:
    - name: _resolution_tag
      dtype: string
    - name: _time_tag
      dtype: string
    - name: dset__artist_occurence_count
      dtype: float64
    - name: quality
      dtype: string
    - name: rating
      dtype: string
    - name: tag_str_artist
      dtype: string
    - name: tag_str_character
      dtype: string
    - name: tag_str_copyright
      dtype: string
    - name: tag_str_general
      dtype: string
    - name: tag_str_meta
      dtype: string
  - name: twi_logfav_convbase__score
    dtype: float64
  - name: edge_density
    dtype: float64
  - name: focus_measure
    dtype: float64
  - name: texture_score
    dtype: float64
  - name: noise_level
    dtype: float64
  - name: noise_level_lap
    dtype: float64
  - name: saturation
    dtype: float64
  - name: contrast
    dtype: float64
  - name: brightness
    dtype: float64
  - name: avg_dynamic_range
    dtype: float64
  - name: filter1_kept
    dtype: bool
  - name: filter2_kept
    dtype: bool
  - name: filter3_kept
    dtype: bool
  - name: openpose__max_total_parts
    dtype: float64
  - name: openpose__max_parts_score
    dtype: float64
  splits:
  - name: train
    num_bytes: 112153557
    num_examples: 55528
  download_size: 45444791
  dataset_size: 112153557
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---


A sample dataset for training various image classifiers with weak supervision.

data curated from full pixiv/danbooru/twitter of over 22M pool