README.md

metadata

datasets:
  - enyoukai/libricount
library_name: datasets
tags:
  - audio
  - speech
dataset_info:
  features:
    - name: mixture_filename
      dtype: string
    - name: mixture_path
      dtype: string
    - name: k
      dtype: int64
    - name: components
      list:
        - name: chosen_start_sec
          dtype: float64
        - name: speaker_id
          dtype: string
        - name: utterance_id
          dtype: string
        - name: utterance_length_sec
          dtype: float64
    - name: audio
      dtype: audio
  splits:
    - name: k1
      num_bytes: 640430247
      num_examples: 2000
    - name: k2
      num_bytes: 640516092
      num_examples: 2000
    - name: k3
      num_bytes: 640602191
      num_examples: 2000
    - name: k4
      num_bytes: 640687924
      num_examples: 2000
    - name: k5
      num_bytes: 640773885
      num_examples: 2000
    - name: k6
      num_bytes: 640859789
      num_examples: 2000
    - name: k7
      num_bytes: 640945405
      num_examples: 2000
    - name: k8
      num_bytes: 641031873
      num_examples: 2000
    - name: k9
      num_bytes: 641117546
      num_examples: 2000
    - name: k10
      num_bytes: 641207541
      num_examples: 2000
  download_size: 6392980636
  dataset_size: 6408172493
configs:
  - config_name: default
    data_files:
      - split: k1
        path: data/k1-*
      - split: k2
        path: data/k2-*
      - split: k3
        path: data/k3-*
      - split: k4
        path: data/k4-*
      - split: k5
        path: data/k5-*
      - split: k6
        path: data/k6-*
      - split: k7
        path: data/k7-*
      - split: k8
        path: data/k8-*
      - split: k9
        path: data/k9-*
      - split: k10
        path: data/k10-*

Libricount

Mixture audios grouped by number of speakers K with per-speaker components and chosen start times. Splits are named 'k1', 'k5', 'k10', etc.

Loading

from datasets import load_dataset

ds = load_dataset("enyoukai/libricount")
print(ds)  # DatasetDict with splits by K (e.g., 'k1', 'k5', 'k10')

k5 = load_dataset("enyoukai/libricount", split="k5")
row = k5[0]
audio = row["audio"]  # dict with path/array

Features

• mixture_filename: str
• mixture_path: str
• k: int (number of speakers in mixture)
• components: list of speaker component dicts (length == k)
• audio: audio column