YOLO-Formatted Dataset of Bottles in Diverse Backgrounds (plastic_bottle & tetra_pak)

The dataset consists of 548 bottle images that are gathered in different real-life scenarios such as streets, parks, shops, and indoor scenes. The pictures were taken in a variety of conditions like variation in lighting, background and object positions to create more variability. All of the images are annotated adequately in the YOLO format, where there are bounding boxes denoting two classes (plastic_bottle and tetra_pak). The annotation files are in the YOLO TXT format (class_id x_center y_center width height with normalized coordinates) and accompanying the images so that it is ready-to-use.

The dataset can help serve the study of computer application in the field of bottle type detection, waste classification, recycling, and environmental monitoring. It can be employed to train and test object detection models to differentiate recycling objects in complicated and noisy backgrounds. There is no predetermined division, and scientists can make their own train, validation, and test subsets as per their requirements.

It should be of particular interest to students, researchers and practitioners who work with a YOLO-based pipeline (e.g., YOLOv5 including YOLOv8) but it can be converted into other formats like Pascal VOC or COCO. By presenting a diverse and labeled set of bottles, this work will help to drive the field of automated waste recognition and reinforce the use of sustainable tools via machine learning.