dontcry / README.md
dontcryai's picture
Upload 5 files
158d507 verified
metadata
license: mit
tags:
  - audio-classification
  - baby-cry-detection
  - pytorch
  - hubert
language:
  - en
library_name: transformers
pipeline_tag: audio-classification

Baby Cry Classification Model

Classifies baby cries into 5 categories to help parents understand their baby's needs.

Categories

  • 🍼 Hungry: Baby needs feeding
  • 😴 Tired: Baby needs sleep
  • 😣 Belly Pain: Digestive discomfort
  • 💨 Burping: Needs burping
  • 😢 Discomfort: General discomfort (diaper, temperature, etc.)

Model Details

  • Base Model: HuBERT (facebook/hubert-base-ls960)
  • Architecture: HuBERT + Custom Classification Head
  • Sample Rate: 16kHz
  • Input Duration: 5 seconds
  • Framework: PyTorch + Transformers

Usage

import requests
import numpy as np

# Load audio file
audio_data = open("baby_cry.wav", "rb").read()

# Call API
API_URL = "https://api-inference.huggingface.co/models/dontcryai/baby-cry-classifier"
headers = {"Authorization": "Bearer YOUR_HF_TOKEN"}

response = requests.post(API_URL, headers=headers, data=audio_data)
result = response.json()

print(result)
# [{'label': 'hungry', 'score': 0.85}, ...]

Training

Trained on custom baby cry dataset with data augmentation.

License

MIT License - Free for commercial and personal use