The Beach Plastic Litter Dataset version 2: BePLi Dataset v2 is an updated version of BePLi Dataset v1. The BePLi Dataset v2 comprises 3722 original beach plastic litter images and corresponding Instance segmentation annotations that were manually annotated. The annotations are provided as MS COCO (Microsoft Common Objects in Context) Object Detection Task formatted JSON file. Plastic litter object classes are defined as 13 classes: "pet_bottle", "other_bottle", "plastic_bag", "box_shaped_case", "other_container", "rope", "other_string", "fishing_net", "buoy", "other_fishing_gear", "styrene_foam", "others", "fragment". All target objects are made from plastics.  This dataset employed RLE encoding for the instance segmentation for a single object and all the annotations are given for single objects.