Research on chainsaw detection using IoT device and machine learning

Project Brief

Jagawana is a wide sensor network system deployed in the forests to prevent ilegal logging. We use sensors to pick up voices in the forests to monitor what happened in the forest in real-time. We use machine learning to process the sounds taken by the sensor and to identify the sounds into various categories, such as chainsaws, trucks, gunshot, and burning sounds. We will be using Android App to monitor and notify the user if suspicious events were happening in the forest.

This project combines machine learning, internet of things, cloud computing, and android application. This project is part of Bangkit Academy's Capstone Project. My role in this project involves creating the Machine Learning Model, designing and developing the Google Cloud Architecture, making the IoT prototype, and creating design and illustration.

The machine learning model is developed on Kaggle using ESC-50 Audio Dataset, Urbansound8k Dataset, and Google's Audioset.

We are using ESP32 and Mosquitto Broker to prototype the working device. The Google Cloud Platform then will receive and store the audio data using Pub/Sub as a trigger for the cloud functions to store the data to Cloud Storage and BigQuery. Every audio data inputted will be processed by our machine learning model deployed on the AI Platform. You can see the project overview on image below.

Jagawana System Overview

Photos

More about Jagawana

Check out the detailed blog post of the project here