Computer Vision Annotation for Autonomous Navigation in Robotics
The primary goal of this project was to develop a high-accuracy object detection model to enhance the safety and navigational efficiency of autonomous robotic mowers operating in complex solar field environments. My responsibilities included analyzing and annotating a large dataset of several thousand images captured from the robots' onboard cameras. I performed detailed labeling tasks, using bounding boxes and polygons to precisely identify and classify a wide range of static and dynamic obstacles, such as solar panel infrastructure, terrain variations, unexpected debris, and vegetation. A rigorous quality assurance process was followed, including peer reviews and iterative refinement of labeling guidelines to ensure maximum consistency and accuracy for training the computer vision model.