Field-Acquired RGB-Depth Image Dataset for Baby Broccoli Detection and Size Estimation Under Varying Illumination Conditions

This dataset contains 1,759 paired RGB images (640×480 pixels) and corresponding 16-bit depth frames of baby broccoli plants acquired in commercial fields using an Intel RealSense D435 camera under daytime and night-time illumination. The images are organised into daytime and night-time folders with matching RGB-depth pairs. The data support research on agricultural computer vision and robotic harvesting, including broccoli detection, segmentation, size estimation, and illumination-robust perception in field environments. The annotated_dimensions_validation_only.zip folder contains ground truth diameter measurements for validation and testing purposes only; the sample size (94 measurements) is not sufficient for training deep learning regression models.