AquaCam: An ML-Enhanced Low-Cost, Deploy-Anywhere Water Level Detection Sensor

Flood detection is problematic, especially in rural areas with limited or nonexistent monitoring infrastructure. Smaller streams and flood-prone areas frequently go unmonitored, leaving these regions vulnerable to flooding. Commercial flood detection systems are expensive and proprietary, making them inaccessible to many communities.In this work, we present AquaCam—a low-cost, open-source flood detection system that uses a Raspberry Pi and a camera to measure the water level of streams automatically. AquaCam autonomously captures data and trains a lightweight convolutional neural network (YOLOv8) using the collected images. This training enables the model to accurately recognize water against natural backgrounds and measure water height.Could AquaCam adapt to new environments? To find out, we tested our trained model at a completely different site without any additional retraining. The results were promising—despite the location change, the system accurately identified water levels, proving that our approach is practical and generalizable. With AquaCam, we take a step toward democratizing flood detection, making it more affordable, accessible, and adaptable for the communities that need it most.