Machine Learning for Image Velocimetry¶
Collaborative research applying deep learning to advance the state-of-the-art of image-based flow measurement — automating and improving the image velocimetry workflow with physics-guided deep learning.
Auto-STIV¶
Auto-STIV is a deep learning framework for fully autonomous two-dimensional streamflow velocity vector field estimation. Developed in collaboration with Pennsylvania State University (Dr. Xiaofeng Liu, Dr. Roberto Fernández, and Dr. Xiaofeng Liu).
Key Innovation¶
Auto-STIV eliminates the need for manual parameter tuning in space-time image velocimetry by using deep learning to automatically detect and extract velocity information from video sequences. This represents a step toward fully autonomous, real-time discharge measurement from camera imagery.
Publication¶
Tenorio, A., Umarova, A., Fernández, R., Engel, F.L., and Liu, X., in review, Auto-STIV — A deep learning framework for fully autonomous two-dimensional streamflow velocity vector field estimation: Water Resources Research.
AIPIV¶
AIPIV (Entropy-guided deep feature particle image velocimetry) applies entropy-guided deep learning for robust river surface velocity estimation. Developed in collaboration with the Stevens Institute of Technology (Dr. Mahmoud Ayyad).
Key Innovation¶
AIPIV integrates information-theory principles (entropy) with deep feature extraction to produce more robust velocity estimates under challenging field conditions — low contrast, variable lighting, and sparse surface tracers.
Publication¶
Ayyad, M., Engel, F.L., Temimi, M., and Henein, M.M.R., in review, AIPIV — Entropy-guided deep feature particle image velocimetry for robust river surface velocity estimation: Water Resources Research.
My Role¶
I serve as co-PI and domain expert on both collaborations, providing:
- Operational image velocimetry expertise and validation datasets
- USGS field data and camera imagery
- Integration requirements for operational deployment
- Advising graduate students conducting the ML research
Collaborators¶
- Pennsylvania State University — Dr. Xiaofeng Liu, Dr. Roberto Fernández, Alejandro. Tenorio (Ph.D. student)
- Stevens Institute of Technology — Dr. Mahmoud Ayyad, Dr. Marouane Temimi
- University of Arizona — Dr. Jennifer Duan
Related¶
- IVy Tools — the operational framework these methods aim to enhance
- ECHO AI Skunkworks — evaluating ML approaches for production readiness