Argus: Detect VOCs with ML and Chemical Engineering

The Problem:

Workers in garment factories (India) are regularly exposed to VOCs

Regular sensors cost > $1000. They're bulky and high-maintenance.

My Solution:

I used dyes called Polydiacetylenes (PDAs), which change color in response to chemical exposure.

By playing with the tail lengths and functional groups, I made an array of PDAs that exhibits a unique response to various VOCs.

I created an smartphone app (with Flutter) that takes a picture of the patch and:

1. Grabs the color values of the dots

2. Normalizes the dots' values against a white background, to control for lighting conditions

3. Uses Principal Component Analysis (PCA) to simplify a 12-direction vector (R, G, and B for 4 dots) into R3.

4. Finds the closest color profile, and determines that VOC.

Users can take advantage of the data with the following features:

1. Offers the ability to directly share on Social Media

2. Logs the data over time to aid with medical diagnoses

3. Has a built-in template email to India's version of OSHA, with the data included.