장르 레이블은 추천 시스템, 자동 생성 플레이리스트, 음악 발견 기능이 돌아가는 기반입니다. 이 글에서는 오디오 미리듣기와 중첩 장르 트리를 활용해 Potato에서 장르 태깅 인터페이스를 구축하는 방법을 보여 드립니다.

기본 장르 분류

yaml

annotation_task_name: "Music Genre Classification"
 
data_files:
  - data/tracks.json
 
item_properties:
  audio_path: audio_preview
  metadata_fields: [artist, title, album]
 
# Show track metadata
display:
  show_metadata: true
  metadata_template: "{{artist}} - {{title}}"
 
annotation_schemes:
  - annotation_type: audio_annotation
    audio_display: waveform
    waveform_color: "#EC4899"
    progress_color: "#F472B6"
    speed_control: true
    show_duration: true
    name: primary_genre
    description: "Select the primary genre"
    labels:
      - Rock
      - Pop
      - Hip-Hop
      - Electronic
      - Jazz
      - Classical
      - R&B
      - Country
      - Metal
      - Folk

중첩 장르 분류 체계

yaml

annotation_schemes:
  # Top-level genre
  - annotation_type: radio
    name: genre
    description: "Primary genre"
    labels:
      - Rock
      - Electronic
      - Hip-Hop
      - Pop
      - Jazz
      - Classical
 
  # Subgenre (conditional)
  - annotation_type: radio
    name: subgenre
    description: "Subgenre"
    conditional:
      depends_on: genre
      options:
        Rock:
          - Classic Rock
          - Alternative
          - Indie Rock
          - Hard Rock
          - Punk
          - Grunge
        Electronic:
          - House
          - Techno
          - Trance
          - Drum & Bass
          - Ambient
          - Dubstep
        "Hip-Hop":
          - East Coast
          - West Coast
          - Trap
          - Conscious
          - Lo-fi
        Pop:
          - Synth Pop
          - Dance Pop
          - Indie Pop
          - K-Pop
          - Teen Pop
        Jazz:
          - Bebop
          - Smooth Jazz
          - Fusion
          - Big Band
          - Cool Jazz
        Classical:
          - Baroque
          - Romantic
          - Contemporary
          - Opera
          - Chamber

다중 레이블 장르 태깅

많은 트랙이 둘 이상의 장르에 걸쳐 있으므로, 어노테이터가 여러 개를 고를 수 있게 하십시오:

yaml

annotation_schemes:
  - annotation_type: multiselect
    name: genres
    description: "Select ALL applicable genres (up to 3)"
    labels:
      - Rock
      - Pop
      - Electronic
      - Hip-Hop
      - R&B
      - Jazz
      - Classical
      - Country
      - Folk
      - Metal
      - Punk
      - Indie
      - World
      - Latin
    min_selections: 1
    max_selections: 3
 
  - annotation_type: radio
    name: primary_genre
    description: "Which is the PRIMARY genre?"
    dynamic_from: genres  # Only show selected genres as options

완전한 음악 어노테이션 설정

yaml

annotation_task_name: "Music Tagging and Classification"
 
data_files:
  - data/music_library.json
 
item_properties:
  audio_path: preview_url
  metadata_fields:
    - artist
    - title
    - album
    - year
    - duration
 
display:
  show_metadata: true
  metadata_layout: card
  fields:
    - label: "Artist"
      field: artist
    - label: "Track"
      field: title
    - label: "Album"
      field: album
    - label: "Year"
      field: year
 
annotation_schemes:
  # Audio playback configuration
  - annotation_type: audio_annotation
    audio_display: waveform
    waveform_color: "#8B5CF6"
    progress_color: "#A78BFA"
    cursor_color: "#F59E0B"
    height: 100
    show_duration: true
    show_current_time: true
    speed_control: true
    speed_options: [0.75, 1.0, 1.25, 1.5]
    volume_control: true
    default_volume: 0.7
    autoplay: false
    loop: true
 
  # Genre classification
  - annotation_type: multiselect
    name: genres
    description: "Select all applicable genres"
    labels:
      - Rock
      - Pop
      - Electronic/Dance
      - Hip-Hop/Rap
      - R&B/Soul
      - Jazz
      - Classical
      - Country
      - Folk/Acoustic
      - Metal
      - Punk
      - Indie
      - World/International
      - Latin
      - Blues
      - Reggae
    min_selections: 1
    max_selections: 3
 
  # Mood/Energy
  - annotation_type: multiselect
    name: mood
    description: "Select the mood(s) of this track"
    labels:
      - Happy/Uplifting
      - Sad/Melancholic
      - Energetic
      - Calm/Relaxing
      - Aggressive
      - Romantic
      - Dark/Moody
      - Nostalgic
      - Motivational
    min_selections: 1
 
  # Energy level
  - annotation_type: likert
    name: energy
    description: "Energy level"
    size: 5
    min_label: "Very low energy"
    max_label: "Very high energy"
 
  # Danceability
  - annotation_type: likert
    name: danceability
    description: "How danceable is this track?"
    size: 5
    min_label: "Not danceable"
    max_label: "Very danceable"
 
  # Vocal content
  - annotation_type: radio
    name: vocals
    description: "Vocal content"
    labels:
      - Instrumental only
      - Mostly instrumental
      - Balanced
      - Mostly vocals
      - Vocals dominant
 
  # Era/Style
  - annotation_type: radio
    name: era
    description: "Musical era/style"
    labels:
      - Pre-1970s (Classic)
      - 1970s
      - 1980s
      - 1990s
      - 2000s
      - 2010s
      - 2020s/Current
 
  # Quality flags
  - annotation_type: multiselect
    name: flags
    description: "Special flags (if applicable)"
    labels:
      - Explicit content
      - Live recording
      - Remix/Cover
      - Holiday/Seasonal
      - Soundtrack
      - Spoken word sections
    required: false
 
  # Confidence
  - annotation_type: likert
    name: confidence
    description: "Confidence in your genre classification"
    size: 5
    min_label: "Uncertain"
    max_label: "Very confident"
 
annotation_guidelines:
  title: "Music Tagging Guidelines"
  content: |
    ## Genre Selection
    - Choose 1-3 genres that best describe the track
    - Select the most specific applicable genre
    - When unsure, lean toward broader categories
 
    ## Mood Assessment
    - Consider the overall emotional tone
    - A track can have multiple moods
    - Think about how the music makes you feel
 
    ## Energy & Danceability
    - Energy: tempo, intensity, power
    - Danceability: rhythm, beat clarity, groove
 
    ## Listening Tips
    - Listen to at least 30 seconds
    - Pay attention to drops, changes, and transitions
    - Consider instrumentation and production style

출력 형식

json

{
  "id": "track_001",
  "audio_url": "/audio/preview_001.mp3",
  "metadata": {
    "artist": "The Beatles",
    "title": "Here Comes the Sun",
    "album": "Abbey Road",
    "year": 1969
  },
  "annotations": {
    "genres": ["Rock", "Pop"],
    "mood": ["Happy/Uplifting", "Nostalgic"],
    "energy": 3,
    "danceability": 2,
    "vocals": "Balanced",
    "era": "Pre-1970s (Classic)",
    "flags": [],
    "confidence": 5
  }
}

음악 어노테이션 팁

좋은 헤드폰이 중요합니다. 노트북 스피커로는 장르 단서를 놓치게 됩니다.
볼륨을 일정하게 유지해 계속 조정하면서 피로해지지 않도록 하십시오.
도입부만 듣지 말고, 변화를 포착할 만큼 각 미리듣기를 충분히 들으십시오.
시작하기 전에 각 장르가 실제로 무엇을 의미하는지 다시 익혀 두십시오.
휴식을 취하십시오. 지친 귀는 더 나쁜 판단을 내립니다.

다음으로 갈 곳

악기를 식별하려면 오디오 이벤트 탐지를 추가하십시오
데이터에 더 다양한 귀를 모으려면 크라우드소싱을 설정하십시오
음악에 대한 일치도 지표를 알아보십시오

오디오 어노테이션 옵션 전체(파형 표시, 재생 컨트롤, 구간 선택)는 원본 문서를 참고하십시오.

전체 오디오 문서는 /docs/features/audio-annotation에 있습니다.