Grooming Detection using Neural Networks and API

This project provides a multimodal system for detecting grooming behavior using both voice and text inputs. It includes a REST API built in Python, running on a virtual environment and deployable via SLURM for GPU acceleration.

Project Structure

final/
├── creating_venv.bash              # Script to create a Python virtual environment
├── requirements.txt                # Python dependencies
├── venv.bash                       # Script to activate the environment
├── API/
│   ├── main.py                     # Entry point for REST API
│   ├── check_val.py                # Validation helper
│   ├── combined_analysis.py        # Core logic for voice+text inference
│   ├── sample.mp3                  # Example audio input
│   ├── run_api.slurm               # SLURM script for API deployment
│   ├── test_combined_analysis.py   # Unit test for analysis logic
│   └── test_endpoints.py           # API endpoint tests

Getting Started

1. Clone the Repository

git clone https://github.com/zayn303/grooming-detection-api.git
cd grooming-detection-api/final

2. Set Up Virtual Environment

bash creating_venv.bash
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Run the API Locally

cd API
python3 main.py

The API will start on http://127.0.0.1:5000/.

5. Run Tests

pytest test_endpoints.py
pytest test_combined_analysis.py

Sample Usage

Use the provided sample.mp3 file to test the grooming detection endpoint once the API is running.

POST /analyze
Content-Type: multipart/form-data
Payload:
- audio: sample.mp3
- transcript: "example conversation text"

SLURM Deployment

For running on a SLURM-managed GPU cluster:

sbatch run_api.slurm

Make sure the cluster has access to the needed Python environment and GPU.

Dependencies

Key libraries include:

• Flask – for REST API
• torch, transformers – for running models
• librosa, torchaudio – for audio preprocessing
• pytest – for testing

Full list in requirements.txt.

Authors

• Andrii Kolomiiets
  Bachelor's thesis, Technical University of Košice

License

This project is licensed under the MIT License.