Deep learning improves droplet detection on challenging condensation surfaces
Analyzing Condensation on Patterned Surfaces via Deep Learning
Machine Learning
Summary
Detecting tiny water droplets on uneven or patterned surfaces is important for studying condensation, which helps power and water systems work better. Existing methods can't find droplets well on such tricky surfaces. The authors created TileSAM3, a tool that uses smart image slicing and AI to find droplets more accurately, even in poor lighting and on rough surfaces. They tested it on two large droplet picture sets and showed it works much better than previous methods, enabling better understanding of how droplets form and behave.
What this means in practice
- •For materials engineers: Analyze how droplets form and behave on textured surfaces to improve condensation materials under realistic conditions.
- •For video analytics developers: Enhance detection of small objects in videos with poor lighting or uneven backgrounds using hierarchical tiling and segmentation.
Authors
Tae E. Cooper, Michael Gee, Sheila Whitman, Jemma Dengel, Anthony Katselas, Duygu Dengiz, Frank Keutsch, Joanna Aizenberg
Abstract
Dropwise condensation is key to power and water systems. State-of-the-art condensation surfaces are often inhomogeneous, either patterned by design or random defects. Inhomogeneous surface features dictate the rates of each stage of dropwise condensation-nucleation, growth, coalescence, and shedding. Characterizing these rates requires the precise detection of thousands of droplets. Yet, existing segmentation methods are unable to detect droplets on inhomogeneous surfaces. We developed TileSAM3, a universal droplet segmentation tool built on the Segment Anything Model 3 (SAM3). Hierarchical tiling with per-tile inference followed by post-processing extended SAM3's detection cap and raised the effective resolution of small droplets. We created two datasets: High Quality Droplet Dataset (N=365,826; HQDD) with homogeneous surfaces and Low Quality Droplet Dataset (N=63,577; LQDD) with patterned, defect-laden surfaces under adverse lighting. TileSAM3 closely reproduced thresholding annotations on HQDD, surpassed it over sixfold on LQDD (Average Precision 0.57 versus 0.088), and outperformed other deep learning models on all metrics evaluated. Combined with KD-tree tracking, TileSAM3 identified localized nucleation and coalescence in narrow active bands on a stiffness-patterned surface. TileSAM3 enables analysis of understudied inhomogeneous substrates with poor lighting and videography which facilitates faster, lower cost research of condensation materials.