Machine Learning Being Used to Identify Sick Marijuana Plants


Well-Known Member
Sep 19, 2009
Reaction score
Mariposa County CA
From MakeTechEasier

Machine Learning Being Used to Identify Sick Marijuana Plants

Laura Tucker 13th Aug 2018

It may no longer be as illegal as it once was, but marijuana/cannabis is still a big business, whether it’s being used recreationally or medically. Because of that, marijuana growers need to know if their plants are sick. Leave it to a researcher to figure out a way to identify sick marijuana plants.

The Birth of a New Technology

A group of hackers and security researchers were exploring “DIY cannabis tech” at DEF CON’s Cannabis Village. Inspired by the computer scientists at Stanford who created a way for artificial intelligence to identify skin cancer, Harry Moreno, a researcher, developed

This website will allow you to upload a picture of your cannabis plant and get a score back that relates to the health of the plant. “Chronic sickness is a project to create a human-level diagnosis tool for cannabis plants,” reads the site.

Medical marijuana is of course used by those with terminal and chronic illnesses as well as those who are dealing with chronic pain. While it’s sometimes used by cancer survivors to help them deal with the effects of chemotherapy, it’s also been used by some to cure their cancer.

So the importance of whether a marijuana plant is healthy or not could be great. We’re not just talking about the loss of healthy plants preventing people from getting high – we’re talking about it preventing people from getting well.

The Future of Identifying Sick Cannabis

Moreno told the others exploring “DIY cannabis tech” that he thinks they could develop this machine learning process even further to have it determine with greater accuracy the health of a marijuana plant.

Currently, the technology used at Chronic Sickness only works with eighty percent accuracy. But the researcher believes this is because of the small size of the training dataset. He started this project with just 3,000 images.

Mashable used a photo with the tool at Chronic Sickness. They received the news that the plant is indeed healthy, and this claim is backed up with a .81 confidence level.

He’s hoping that in the future this technology will be able to identify that a plant is sick while also naming the illness in question.

“Let’s make a free predictive model for cannabis disease,” he suggested. And to get started, he’s looking for more cannabis photos, ones that have already been labeled.

Taking It Even Further

But this could go much further than being used for weed. Remember that Moreno was inspired by the technology that identifies skin cancer. What other health information could technology answer?

Could it be used to spot some other illnesses? Maybe instead of always having to go to a doctor, we can make this process more remote. Maybe we can just send the doctor a picture of what is ailing us.

The doctor could use the technology, identify the illness, and send pharmaceuticals to fix it or refer the patient to a specialist. Of course it wouldn’t work for all illnesses, but it could work with illnesses that have some physical characteristic to observe such as shingles.



Well-Known Member
Apr 18, 2014
Reaction score
I used to go to DEF CON every year. Amazing stuff. I have not been for about a decade but it looks like they added 'cannabis village'. LOL

Looks like I'll be heading to Vegas next year. LOL