Meet Agrobot, an autonomous robotic harvester that’s poised to revolutionize the agriculture industry. Agrobot works with the world’s leading farmers to automate berry harvesting through the power of artificial intelligence.

Agrobot uses deep learning to determine when to pick fruit at its ripest. Up to 24 robotic arms grip and cut the fruit from its stem to meet the farmer’s quality standards. Agrobot uses a 3D sensing scanner with short-range integrated color and infrared depth sensors to capture the details and identify when fruit is ready for the picking.

Agrobot, powered by NVIDIA Jetson, currently assesses individual strawberries for harvesting using semantic segmentation. This method relies on a convolutional neural network to classify each pixel of the incoming image as belonging to the ‘strawberry’ class or not. However, if strawberries overlap in the image, then semantic segmentation isn’t enough. Agrobot is looking at extending their system to perform instance segmentation where each strawberry is unique.

A bounding box is generated for each strawberry by a neural network that performs object detection. The system uses the unique bounding boxes to generate instance segmentation. The areas where bounding boxes overlap can be further refined by weighing every pixel against the likelihood of it belonging to an overlapping instance. The result allows the system to more accurately assess individual strawberries for harvesting, which will cut down costs and lead to less wasted fruit.

Agriculture will continue to evolve through the help of AI and automating processes will have interesting and useful benefits to society, lowering costs and helping farmers reap the fruits of their labor all year round.

See Agrobot in action in episode 9 of the ‘I am AI’ Docuseries below:

Via Forbes