CauliFCount-BD: Open-Field Cauliflower Detection and Counting Dataset

CauliFCount-BD is an open-field cauliflower image dataset developed for reproductive structure detection and counting in precision agriculture. The images were collected from four agricultural sites in Bogura and Dhaka districts, Bangladesh, during the winter cultivation season from November 2025 to January 2026.

Images were captured at the cauliflower curd formation stage, when plants were approximately 52 to 65 days old, under natural morning, afternoon, and evening field conditions. The dataset includes smartphone images captured using Google Pixel 7 and Redmi Note 10S devices, with resolutions of 3072 × 4080, 2464 × 3280, and 3472 × 4624 pixels.

The dataset contains 7,683 manually annotated cauliflower instances, divided into training, validation, and test subsets with 6,144, 760, and 779 annotated instances, respectively. Annotations were prepared using the VGG Image Annotator (VIA) and verified through manual inspection.

This dataset supports research in crop monitoring, object detection, reproductive structure counting, plant phenotyping, yield estimation, and real-world precision agriculture.