Beach-O YOLO Bottle Perception Dataset

This dataset supports the evaluation of a YOLO-based visual perception subsystem for bottle detection in a beach-cleaning robotic platform. It includes a curated YOLO-format image dataset with a sequence-aware train/validation split, metadata, checksums, sample preview files, and a RealSense BAG sequence used for offline replay evaluation. The dataset contains 1585 images and 1585 YOLO-format label files, with 3167 annotated bottle instances. The final sequence-aware split includes 1267 training images and 318 validation images. The RealSense BAG file, 20241128_133505.bag, was recorded in controlled sandy terrain and used for replay-based perception evaluation. The archive is distributed in three ZIP files: the YOLO-format dataset, the RealSense BAG sequence, and documentation/metadata/checksum files. The accompanying GitHub repository provides dataset documentation, access instructions, metadata, and checksum manifests. This dataset is associated with the manuscript “YOLO-Based Visual Perception for Bottle Detection in a Beach-Cleaning Robot Using RealSense BAG Sequences and Embedded Deployment Analysis”.