Detecting Weed Through Edge Computer Vision

Impact-driven startup Weedbot is developing a laser weeding machinery for farmers that can localize plants, distinguish between crops and weeds and remove weeds with a laser beam. 50 technology changemakers built machine learning models to facilitate pesticide-free food production.

Impact

The use of laser weeding can have a significant impact by reducing the need for chemical herbicides, leading to a reduction in soil and groundwater pollution. This can enable pesticide-free food production, making organic food more affordable and encouraging people to adopt a healthier lifestyle.

Project outcomes

The team successfully developed a high-speed plant image recognition neural network that meets the necessary input and output requirements for real-time weed segmentation with a precision of 1-2mm. The model has a speed of 12ms per image or faster, achieving recognition precision of 100-110% of crop polygon, which means that up to 10% false positives are allowed. The captured image can cover a 200x200mm working area with carrot seedlings as the object to be recognized. All carrots are recognized by the software, and up to 10% false positives are acceptable. The weed segmentation can be done by the same AI that detects carrots or a separate script like PlantCV. These achievements will facilitate the implementation of laser weeding technology, which can reduce or even eliminate the use of chemical herbicides, leading to pesticide-free food production and promoting healthy lifestyles.

Sample image with annotated carrots: 

Insight from Weedbot CEO on the project