Kindbot Buddy

An AI powered plant diagnosis engine

Make sure to go and test Buddy on your own grow!

Imagine finding a discolored leaf peeking out at you from the grow space. What happened? Perhaps you’d google 'yellow cannabis leaves' or post a photo on the cannabis forums. Maybe you have an experienced friend at the local hydro shop.

Some will suggest an exotic virus or genetics or compare feeding philosophies. Many problems show in similar ways and you’ll need to make sense of multiple theories. For example, your burned leaf edges may indicate heat/light stress or a potassium deficiency.

Often, you'll be referred to scan a deficiency chart like the one below:

Meanwhile, your plant’s health degrades and you're beginning to consider the advice to simply chuck the plant.

Although your quest began with finding out what, if anything, is wrong with your plant, you are forced to consider which of MANY problems looks most like yours.

Buddy was designed to reduce the bias and ambiguity by analyzing your photos, reading the stress signs like an expert would and connecting you to evidence-based recommendations.

How does Buddy work?

Instead of searching the internet for “brown spots cannabis leaf”, we aimed to reproduce how humans make diagnosis with image recognition. Imagine a computer cross-referencing your photo against that deficiency chart.

Let's google search “magnesium cannabis leaf”.

After reviewing a few examples, you’d learn to spot that deficiency based on the yellow-striped leaves.

We took this idea further by collecting images from thousands of growers to learn patterns to identify many common environmental stressors such as light burn or nutrient issues like potassium deficiency or nitrogen excess. We also identify common pests like leaf miners and spider mites and learn to sex the plants with close-up photos of their preflowers.

Buddy narrows your search, based on image content using the state-of-the-art in AI.

Buddy instantly produces the top three most likely diagnoses and resources for further investigation. These results are based on the image content as colors and textures indicate different conditions. Buddy has seen a lot of weed, leveraging the collective diagnosis wisdom scattered across the internet.

Buddy's pattern recognition models have the capability of improvement as the app sees more examples.

What's next?

We're going mobile with an Android/iOS app featuring real time diagnosis. Coming to the Google Play store and App store soon!