There’s plenty of information online to help people figure out whether their plants are healthy and what they need to do to make them better. But they can be distributed far across the Internet and sometimes hard to find, and aren’t necessarily available on demand, according to Martín Capeletto.
That’s why the team, a bunch of plant nuts (though they wouldn’t tell us exactly WHICH plants), decided to build Plantbot. You take or upload a photo of a plant, and using some machine learning tools available online, it’ll tell you whether or not your plant is healthy and what might be wrong with it. The process takes a few seconds, and then if users have additional questions, they can query them a chat interface in the app. The company demoed the app at the TechCrunch Disrupt 2017 Hackathon.
“We are working on services usually, but we do want to launch products between the company, and every time the Hackathon is an opportunity to do that,” Capeletto said. “Folks all over the internet are full of questions about the plants, and I love gardening, love the forums, the questions are always the same.”
There are a couple common problems that the app can diagnose — for example, a differently colored leaf might signal that there’s not enough nitrogen in the soil, and brown tips on the leaves may mean there are too many nutrients in the soil, Capeletto said. This handles the issue of quickly answering questions that most plant owners might have.
The team feeds in photos themselves with some common problems, and then matches it against the photos that are coming in, so there will be some hurdles to getting enough data online and encompassing the wide array of problems. But it should be possible to dig through all the information that’s available online, along with accompanying photos, to build a big enough lexicon to get a good handle of most problems for plants and solutions. And they often require simple fixes, Capeletto said
The team behind Plantbot — an international group consisting of Capeletto, Hernan Berroja Albiz, Agustín Capeletto, and Iago Rodríguez López — aren’t strangers to Hackathons. In fact, this is their sixth hack from TechCrunch Disrupt’s Hackathons. Last time around, the team built an app that would crawl through reviews to surface whether or not you’d like something at a venue or restaurant (like whether a restaurant has good tacos, for example).
This isn’t entirely a new idea, and the concept of diagnosing plant health with a camera has been around for a while. In 2013, a project called Infragram sought to do the same thing with a camera. But smartphone cameras are improving, and so is image recognition and comparing technology. If the team can keep up with putting together enough data — or maybe figure out a way to do it algorithmically — it could end up being something pretty useful for plant owners around the world.