Water Detection Service

This service aims to detect malfunctioning irrigation systems using footage acquired by a drone. A machine learning-based algorithm detects the water spray using a Mask Regional Convolutional Neural Network (MRCNN)¹. The service is available through a Streamlit web app.

1 - C. K. G. Albuquerque, S. Polimante, A. Torre-Neto and R. C. Prati, "Water spray detection for smart irrigation systems with Mask R-CNN and UAV footage," 2020 IEEE International Workshop on Metrology for Agriculture and Forestry (MetroAgriFor), Trento, 2020, pp. 236-240, doi: 10.1109/MetroAgriFor50201.2020.9277542. See full article

Installation

Switch to the service branch.

git checkout service

Create the Conda Virtual Environment using the file water_serv.yml located in the env/ directory. (linux-64 platforms)

conda env create -f env/water_serv.yml

Activate the new environment

conda activate water_serv

Usage

Start the Streamlit web app.

streamlit run streamlit_app/app.py

The web app should launch as a new browser tab. If the web app is not launched, access it at the address specified on the terminal. The default address is http://localhost:8501.