An RGB-D time-series dataset of white button mushroom growth for instance segmentation

Abstract

Accurate detection and segmentation of individual mushroom caps in densely populated cultivation beds remain key challenges for robotic harvesting of white button mushrooms (Agaricus bisporus), where precise boundary delineation is required for size estimation, grasp planning, and collision-free manipulation. The data was developed to support research on computer vision-based perception systems for autonomous mushroom harvesting. The dataset comprises of 129 time-series RGB and depth images of mushrooms from a top-view perspective, representing mushroom growth and spatial distribution under an indoor mushroom production environment. Each sample includes an RGB image, a corresponding depth image, and pixel-level segmentation annotations of individual mushrooms in COCO JSON format. The inclusion of depth information enables the extraction of geometric features such as cap height, curvature, and relative positioning between neighboring mushrooms. This dataset provides a resource for training and benchmarking algorithms in instance segmentation, growth analysis, and perception for agricultural robotics, with particular relevance to automated harvesting and RGB–D scene understanding in complex, densely populated agricultural settings.