DripTect
On-ground sensors based flood detection system.
Created on 24th August 2025
•
DripTect
On-ground sensors based flood detection system.
The problem DripTect solves
🌊 The Problem It Solves
Traditional flood alerts mostly rely on broad weather forecasts, which often fail to provide ground-level accuracy. They generate generic warnings that may be too late, too wide in scope, or irrelevant for the communities actually at risk. This lack of precision leaves vulnerable populations, farmers, and cities unprepared for sudden floods.
DripTect changes this by:
-
Fusing on-ground sensor data with forecasts – Unlike one-dimensional weather alerts, DripTect combines real-time readings from soil, water, rainfall, humidity, and temperature sensors with predictive models for richer insights.
-
Delivering localized risk levels – Instead of blanket warnings, it provides region-specific flood classifications (Safe / Risk / Flood), enabling actionable, hyper-local decisions.
-
Enabling low-cost, scalable deployment – Built on ESP32 and Arduino, DripTect is affordable and energy-efficient, making it accessible even for rural, low-resource flood-prone areas.
-
Offering a multi-dimensional risk model – Goes beyond rainfall to monitor soil moisture, humidity, wind speed, water levels, and temperature, capturing the true complexity of floods.
-
Improving continuously – Its machine learning models learn from every new dataset, adapting to climate changes and increasing accuracy season after season.
👉 With DripTect, flood management evolves from broad, reactive alerts to precise, proactive intelligence, ensuring safer communities and smarter disaster response.
Challenges we ran into
âš¡ Challenges I Ran Into
1. Hardware–Software Integration
One major hurdle was synchronizing the ESP32 hardware sensors with the backend pipeline. Sensor data often came in noisy, inconsistent formats (e.g., fluctuating soil moisture readings due to temperature drift).
✅ Solution: I implemented data smoothing & calibration techniques and wrote a preprocessing layer to filter anomalies before feeding the ML models.
2. Dataset Limitations
Flood datasets combining real hydrological parameters with sensor-based simulations were scarce. Training a model directly on limited lab data caused overfitting issues.
✅ Solution: I combined open-source flood/weather datasets (IMD, NOAA, Kaggle) with synthetic sensor data, and used data augmentation & hypertuning to improve model generalization.
Tracks Applied (2)