This page presents the computational imaging results for analyzing aerosol reactivity outcomes from our automated synthesis experiments.
Microscope images are analyzed in RGB, HSV, and LAB color spaces to understand droplet formation and chemical reactions. Each color space provides different insights into the reaction outcomes.
Droplet detection using Facebook's SAM (Segment Anything Model) with filtering by size (36-1500 pixels) and circularity (>0.8) to identify valid microdroplets.
Representative colors extracted from each detected droplet using various methods including center pixel, maximum saturation, and other color space techniques.
Histogram showing the diameter distribution of detected droplets, calculated from segmented areas assuming circular geometry.
Pairplot visualization of RGB color distributions with fitted 3-component Gaussian mixture models to classify reaction outcomes.
Individual droplets plotted on a triangular simplex based on their probability of belonging to each of the three identified clusters, with separate views for each cluster.
Summary visualization showing droplet size distributions and counts across all experimental samples, organized by experiment type.
Analysis of the uncertainty in droplet cluster assignments, showing the confidence of the statistical model in classifying each droplet into reaction outcome categories.