From eyes to cameras to kinetics : computer vision for chemical process monitoring

Tools

Barrington, Henry and Donnachie, Kristin and Fyfe, Calum and McCabe, Timothy J.D. and Ward, Cameron and Reid, Marc (2026) From eyes to cameras to kinetics : computer vision for chemical process monitoring. Accounts of Chemical Research, 59 (11). pp. 1846-1867. ISSN 0001-4842 (https://doi.org/10.1021/acs.accounts.6c00137)

[thumbnail of Barrington-etal-ACR-2026-From-eyes-to-cameras-to-kinetics-computer-vision]
Preview
 Text. Filename: Barrington-etal-ACR-2026-From-eyes-to-cameras-to-kinetics-computer-vision.pdf
Final Published Version
License: Creative Commons Attribution 4.0 logo
Download (5MB)| Preview

Abstract

Chemical manufacturing is dominated by color changes that offer quantifiable insights into a huge variety of transformations. Yet, despite the vast array of chemical phenomena we experience by eye, few reaction monitoring tools are designed to measure what we see. This Account describes our journey developing computer vision to turn any camera into a chemistry-agnostic reaction monitoring tool able to transform everyday visual observations into quantitative kinetic data. Chemists routinely record color changes, precipitation, and mixing patterns without systematically leveraging this information. Our work emerged from an idea inspired by early career resource constraints: if visual changes can be described qualitatively, they can be quantified digitally, and cheaply. This insight, catalyzed by Raspberry Pi camera experiments and a serendipitous industrial need, led us to develop computer vision methods that enable bulk process monitoring across scales. Central to our contributions is the development of Kineticolor, a software platform converting video recordings into time-resolved kinetic information. Beyond color averaging, we developed spatially resolved analyses, providing time-resolved texture metrics to reveal mixing phenomena, even in small scale applications where the temptation is often to ignore mixing effects. Crucially, our methods are scale-agnostic, meaning identical algorithms apply from microscale wells to industrial reactors. Our industry-focused collaborations have highlighted computer vision’s complementary role in process monitoring: while molecular techniques provide chemical identity, noninvasive visual methods capture bulk phenomena missed by point measurements. Our work serves modern automation laboratories looking for their next complementary data stream, as well as enabling accessible monitoring using standard cameras for resource-constrained environments. If chemists observe visual changes in laboratory notebooks, computer vision can quantify those tacit insights. The principle is simple: if you can see it, these methods can measure it.

ORCID iDs

Barrington, Henry ORCID logoORCID: https://orcid.org/0009-0005-2978-670X, Donnachie, Kristin, Fyfe, Calum, McCabe, Timothy J.D. ORCID logoORCID: https://orcid.org/0009-0002-2524-0513, Ward, Cameron and Reid, Marc ORCID logoORCID: https://orcid.org/0000-0003-4394-3132;
CORE (COnnecting REpositories)