The AOC fruit detector is a ROS 2 package for detecting fruit for harvesting. It supports strawberries and tomatoes and uses Mask R-CNN instance segmentation to find each fruit, classify ripeness and estimate its size and 3D position from an RGB-D camera.

Features

• Subscribes to RGB-D camera topics and publishes bounding boxes, segmentation masks, 3D poses, ripeness and size for each detected fruit, plus annotated images and RViz markers.
• Pre-trained models for strawberry and tomato, downloaded automatically on first run.
• A training procedure to refine the models with your own data or train models for other crops.
• Development container and Dockerfile for a consistent setup.

The detector features in a hands-on demonstration of training and deployment at the ROS in the Field workshop, ROSCon UK 2026.

Get it

Source code, installation and usage instructions are in the GitHub repository and its wiki.

Licence: Apache-2.0.