AI for Precision Agriculture

See how drones and neural networks can reduce herbicide use at scale.

Challenge

Did you know that $30 billion are wasted every year during weed-killer application in agriculture? To put that in context, $30 billion is a GDP of a small country. And it’s not just the money - herbicides can be dangerous for the environment as well as for our health. A German agro-tech company SAM DIMENSION had a bold vision to stop this tremendous waste and chose Bohemian AI as a partner for that journey.

So why do farmers need to use so much herbicides when only a small fraction of them is actually useful? In a nutshell, fields are large and weeds are small, and until very recently there was no technology that could find the weeds fast enough and precisely enough, so a farmer would typically spray the entire field to make 100% sure that all weed plants are hit by the weed-killer.

If you wish to learn more about the technical challenges of weed scouting, we have covered them in detail in this blog post.

Solution

The best answer, we believe, is a combination of drones and AI working together to map a field and precisely locate even the smallest weed plants. Any modern sprayer can then use this map to apply weed-killers only where they’re absolutely necessary. Up to 90% herbicides can be saved this way.

To bring the AI component to life, we designed and implemented processes, infrastructure, tools, neural networks, and user interfaces to handle a large amount of high-res images and to find weeds on these images, pixel by pixel.

Looking for something similar?

TALK TO OUR EXPERT

Technology

  • Cloud infrastructure for data storage and AI training on AWS handling terabytes of data
  • End-to-end data pipeline using Python
  • User interface for the data pipeline using Django
  • Proprietary annotation tool using Python and JavaScript
  • Custom neural networks using PyTorch
  • Edge inference optimized for Nvidia Jetson

Examples

Image taken by a drone on a sugar beet field

Weed scouting example input

Neural network's prediction (crop = blue, weeds = red)

Weed scouting example prediction

Sounds interesting?

If you’re interested in learning more about SAM DIMENSION’s AI, or if you have your own computer vision project in mind, let us know. Our pre-sales consultations are free, with no strings attached. Go ahead and schedule a call right away!

Read More Case Studies