---
language: en
tags:
  - audio-classification
  - causal-representation
  - infant-cry-detection
license: mit
datasets:
  - custom-audio-dataset
metrics:
  - event-based-f1
  - iou
  - accuracy
---

# Infant Cry Detection Using Causal Temporal Representation

This model detects infant cries using a novel **causal temporal representation** framework. By integrating causal reasoning into the data-generating process (DGP), the model aims to enhance the interpretability and reliability of cry detection systems. 

## Features
- **Causal Data Generating Process**: Incorporates mathematical causal assumptions to define the relationship between audio features and annotations.
- **Supervised Models**: Includes pre-trained state-of-the-art models:
  - Bidirectional LSTM
  - Transformer
  - MobileNet V2
- **Event-Based Metrics**: Tailored for time-sensitive detection tasks:
  - Event-based F1-score
  - Intersection over Union (IOU)
- **Interactive Example**: Jupyter Notebook with step-by-step usage demonstrations.

![Causal Graph](https://huggingface.co/CAPYLEE/CRSTC/blob/main/casual_graph.png)

---

## How to Use

You can load the model directly from Hugging Face:

```python