About

Winning Video of the 2026 TC105 Student Video Competition | Deep learning-based particle tracking velocimetry (PTV) for spherical and non-spherical particles

Particle-resolved measurements provide a powerful way to understand the mechanics of bedload sediment transport, but conventional particle tracking methods often struggle with dense flows, particle overlap, rotation, and irregular grain shapes. This video introduces a deep learning-based particle tracking velocimetry (PTV) framework that combines YOLO object detection with Kalman-filter-based tracking to identify and follow both spherical and non-spherical particles in complex granular flows. The method is developed and validated using high-speed flume experiments together with synthetic datasets from discrete element method simulations. It enables robust reconstruction of individual particle trajectories and provides particle-scale measurements of translational velocity, rotational velocity, shear rate, and granular temperature. Application to bedload transport reveals systematic differences between spherical and irregular particles and demonstrates how particle shape influences collisional agitation, rotation, and energy dissipation. The framework provides a new experimental tool for linking grain-scale dynamics with macroscopic sediment transport behavior.