Artificial Intelligence

Infrastructure Inspection:
Below are some samples of our “PipeYOLO” predictions on representative sewer CCTV frames, with bounding boxes and defect labels to illustrate the model performance under real inspection conditions (low light, noise, reflections, and textured pipe backgrounds). The model also provides a confidence score with the bounding box, which is a value from 0 to 1 that represents the product of the probability that an object exists in a box and how accurately that box overlaps with the actual object:

Stormwater Monitoring:
With the use of cameras, sensors and artificial intelligence, we are developing a warning system for when the San Diego River gets too high. NBC 7 shows how this new technology could save lives (November 12, 2025): LINK

Weather Forecast , e.g. Downscaling Techniques:
We also use linear mathematical machine learning cores, such as Multiple Linear Regression (MLR), Ridge Regression (RR), Multivariate Adaptive Regression Splines (MARS) and the Model Tree (MT), or non-linear cores, such as cubic-order MARS, k-Nearest Neighbor (kNN) and Genetic Algorithm-optimized Support Vector Machine (GA-SVM) – depending on the problem structure. All these methods provide data partitioning for solving water resources problems.

For example, we have used machine learning to forecast future rainfall, and then have compared the results with a conventional model. Comparison indicates that our machine could significantly improve forecasting efficiency in terms of reproducing standard deviation and skewness for both calibration and validation periods.

Below are the parameters and structure of our Model Tree for forecasting the rainfall in the month of January as an example. x1 to x5 are different meteorological variables.

YouTube Channel (tutorials on SWMM, HEC-RAS, etc.): https://www.youtube.com/@hassandavani2863
LinkedIn Page:
https://www.linkedin.com/in/hassantd/
Google Scholar page:
https://scholar.google.com/citations?user=P9kZ2awAAAAJ&hl=en