Gunshot Detection

I came to America about a month ago. Just two days later, there were reports of mass shootings spread throughout the US. I did some research on American gun violence and found that there is a larger problem underneath. People use guns for suicides much more than to kill others, and only about 15% survive. This inspired gunshot detection

Microphones like the ones in home assistants are ubiquitous and active. The solution uses those microphones to analyze sounds and identify gunshots.

I found an audio dataset containing gunshot sample, amongst others. I ran the data through a Keras sequential NN to get a classifier, to flag gunshot sounds.

The dataset was highly unbalanced, and more samples of gunshots needed to be added. I tried using various possible data-points sourced from a million sources to create a usable dataset which could train a classifier to be accurate

I learnt using Machine Learning APIs like Keras, how to preprocess data, how to encode and extract data from audio files.