Mobile App to predict Water Quality
Eliminates the need for expensive water testing equipment and makes data about lake quality more easily available.
Developed based on the research linking number of bird species at a lake to the water quality at the lake
Predict Lake Water Quality
Using Birdsong Analysis
Features
1. 40 local Indian bird species categorised
2. 4GB of voice samples (downloaded from xeno-canto.org) used for training the CNN model
3. Mel spectrogram with hi-pass filter used to convert .wav files to images
4. Transfer Learning using EfficientNet B3 to train model
5. Training accuracy of 77.6% achieved
Not yet available for public download
A Bit About the App
The key to protecting existing lakes is first documenting them, and then ensuring that this data remains openly accessible to the public. This data can then be use for research, education or conservation. Aboat Time's mobile application provides users with accurate real-time information about the water quality of lakes in their vicinity thus eliminating the need for expensive water testing equipment.
​
Though we identified some apps that identify number of bird species at a locality using bird song analysis and other apps that crowdsource information on water quality of lakes, ours is the first to link the two and use the number of bird species to estimate water quality.
Technovation 2023
Aboat Time is the Asia Regional Winner for their app submission for the 2023 season of Technovation. Our app aims to combat the lack of real time monitoring of water quality by making data about water quality of lakes more accessible. Our trained CNN model is capable of identifying 40 Indian bird species through birdsong analysis.
