A New Startup Wants to Diagnose Sick Cannabis Plants With AI

At this year’s massive hacker conference, DEF CON, one entrepreneur unveiled technology that he says could change the face of cannabis cultivation. Harry Moreno told attendees at the “DIY cannabis tech” seminar in DEF CON’s Cannabis Village that he wants to use machine learning to identify sick cannabis plants, Mashable reports.

The project, called Chronic Sickness, was inspired by Stanford scientists who taught an AI to identify photos of skin cancer. Moreno’s site is similar in that it allows visitors to upload a picture of a sick cannabis plant or a healthy one, and receive a health score.

“Chronic sickness is a project to create a human-level diagnosis tool for Cannabis plants,” his website reads. The tool works by feeding correctly identified images into the AI and telling it which are healthy and which are not. Through repetition, it’s then able to begin making its own distinctions.

Currently, it’s only operating at 80 percent accuracy, Mashable reports, but Moreno expects that number to increase as the database of correctly identified images grows. The AI started with 3,000 images to reference. While the current model can only tell if a plant is sick or not, he hopes that by feeding it more images with more specific labels, it will eventually be able to identify precisely what ails the plant long before it’s too late to save.

Mold and spider mites are two of the biggest threats to a cannabis plant. Experienced growers can pick out the telltale webs of spider mites from a mile away, but perhaps this tool could spot mites before they even get that far along. And identifying mold spots sooner would save growers a headache in the long run. The same goes for russet mites, which are also an issue.

In Moreno’s eyes, AI is the best way to fight those pests before crops are lost. He hopes that growers will eventually be able to snap a photo of a plant on their smartphone, upload it to the site, and know exactly what’s wrong within seconds.

“Let’s make a free predictive model for cannabis disease,” he told the DEF CON crowd. Given a robust enough training dataset, Moreno thinks Chronic Sickness will be the tool needed to identify sick cannabis plants and ultimately save them too.