Scream_Detection_AI-ML_Model

This program uses a deep-neural network to detect the scream and take immediate actions to send help.

Table of Contents

• Introduction
• Features
• Usage
• Results
• Contributing
• License

Introduction

The Scream Detection AI/ML Model is an innovative system developed to enhance the safety and security of our college community. The project aims to detect distress calls and screams in real-time, enabling immediate response and providing assistance to individuals in danger. This repository contains the source code and documentation related to the development and evaluation of the AI/ML model.

Features

• Accurate detection of distress calls and screams
• Real-time response system
• Multi-layer perceptron (MLP) model for classification
• Support vector machines (SVM) for additional confirmation
• Visualizations for performance analysis
• Easy integration with specified locations in the community

Usage

1. Prepare the dataset:

• Collect audio samples of distress calls, screams, and background noise.
• Organize the samples in the resources folder.
• Update the newresources.csv file with the appropriate labels.

2. Train the AI/ML model:

• Run the script to train the multi-layer perceptron model.
• The model will be saved as a file for future use.

3. Evaluate the model:

• Run the script to assess the model's performance.
• The script generates visualizations, including training loss, accuracy, confusion matrix, audio waveform, spectrogram, and precision-recall curve.

4. Integration:

• Install the model in specified locations within the community for real-time scream detection.
• Use the process_file() and svm_process() functions in the code to process audio files and detect distress calls.

Results

The project has demonstrated promising results in accurately detecting distress calls and screams. The AI/ML model achieved high accuracy and has been validated through extensive training and evaluation. The visualizations provide insights into the model's performance, helping assess its effectiveness.

Contributing

Contributions are welcome! If you have any ideas or suggestions to enhance the project, please feel free to open an issue or submit a pull request. Let's collaborate to make our community safer together.

License

This project is licensed under the MIT License.