hazooree / real-time-implementation-of-gunshot-detection-system

Real Time Implementation of Gunshot Detection System

Real Time Implementation of Gunshot Detection System

This is our EE512 Machine Learning semester project at Department of ELectrical Engineering at ITU.

1st: datasets merging, exploration and feature extraction notebook

• Merging Data From Multiple Sources
• Initial Dataset Exploration
• Feature Extraction

2nd: Train Muliple Models notebook

• Extract More Features
• Train Test Split
• APPLY PCA
• Train Models

Trained Models: [GPUs used for Training (Nvidia Geforce GTX 970 and 1070)]:

• Neural Network
• Convolutional Neural Network
• Isolation Forest (Anomaly Detection)
• OneClass SVM (Anomaly Detection)
• Autoencoders (Anomaly Detection)

Supervisor:

Dr. Ali Ahmed

Group Members:

Fawad Arshad ([email protected])

Hazoor Ahmad ([email protected])

Jawad Arshad ([email protected])

Zeeshan Haider ([email protected])

About Repository

Dependencies

Our models need librosa, along with other libraries imported at the top of every ipynb files.

Training Details

You can make your own dataset of audios or use Google, Mivia or Kaggle dataset for gunshot detection. Your can aso download dataset from https://drive.google.com/drive/folders/1_H68GGKBqGsCxQVNZb8lxFy_apRUlhvF?usp=sharing

Working with Trained Models

You can also write your own codes for the evaluation of our Trained Models.

Realtime Simulation Codes

Realtime simulation codes are also included.