README.md

metadata

license: cc-by-4.0
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
dataset_info:
  features:
    - name: text
      dtype: string
    - name: id
      dtype: string
    - name: language
      dtype: string
    - name: is_tts
      dtype: int64
    - name: audio
      dtype:
        audio:
          sampling_rate: 16000
  splits:
    - name: train
      num_bytes: 17420974680.498
      num_examples: 31102
    - name: test
      num_bytes: 1510248128.015
      num_examples: 2635
  download_size: 18318102787
  dataset_size: 18931222808.513

IndicTTS Deepfake Detection Challenge

Participants will use the SherryT997/IndicTTS-Deepfake-Challenge-Data dataset, hosted on Hugging Face. This dataset consists of train and test splits and contains speech samples in 16 Indian languages, along with metadata for each audio clip.

🚀 Dataset to Use: SherryT997/IndicTTS-Deepfake-Challenge-Data

This is the official dataset for the challenge and must be used for training and evaluation.

---

📌 Dataset Structure

The dataset consists of the following columns:

Column	Type	Description
id	int	Unique identifier for each audio sample
audio	Audio()	Audio file in a format supported by Hugging Face's datasets library
text	string	Transcription of the spoken audio
language	string	Language of the audio sample
is_tts	int	Label indicating if the audio is AI-generated (1) or real (0) (only available in train set)

🔹 Important: The test set does not contain the is_tts labels. Instead, for all test samples, is_tts is set to -1, and participants must predict its value.

---

📊 Dataset Statistics

• Train Set: 31,102 samples (contains is_tts labels)
• Test Set: 2,664 samples (is_tts = -1 for all rows)

---

🗣️ Languages Included

This dataset covers 16 Indian languages, ensuring diversity in speech patterns and phonetics:

✅ Assamese
✅ Bengali
✅ Bodo
✅ Dogri
✅ Kannada
✅ Malayalam
✅ Marathi
✅ Sanskrit
✅ Nepali
✅ English
✅ Telugu
✅ Hindi
✅ Odia
✅ Manipuri
✅ Gujarati
✅ Tamil

---

📥 Accessing the Dataset

Participants can load the dataset directly from Hugging Face using the datasets library:

from datasets import load_dataset

# Official dataset for this challenge
dataset = load_dataset("SherryT997/IndicTTS-Deepfake-Challenge-Data")  

# Train and test splits
train_data = dataset["train"]  # Contains 'is_tts' labels
test_data = dataset["test"]  # 'is_tts' is -1 for all rows