KNR – PET –7 : An industrial – grade image dataset of deformed and contaminated PET bottles for fine-grained sorting in emerging markets

The KNR – PET –7 dataset comprises over 5,000 high-resolution RGB images of post-consumer Polyethylene Terephthalate (PET) plastic bottles, captured directly at an operational industrial recycling facility in Vietnam. KNR – PET –7 is specifically designed to reflect the in-the-wild complexities of waste streams in emerging markets, featuring severe morphological deformations (crushed, flattened) and heavy contamination (mud, residual liquids). The dataset is meticulously annotated for object detection tasks using the standard YOLO bounding box format (.txt files). It features 7 fine-grained industrial PET categories: B1, B2.1, B2.2, B3, B4, B5 and B6. This dataset serves as a robust and challenging benchmark for developing real-time machine vision and AI-driven robotic sorting systems in circular economy initiatives.KNR-PET-7数据集包含超过5000张高分辨率RGB图像，素材为消费后聚对苯二甲酸乙二醇酯（Polyethylene Terephthalate，PET）塑料瓶，直接采集自越南某运营中的工业回收工厂。该数据集专为反映新兴市场废物流的真实场景复杂性而打造，样本普遍存在严重形态变形（压碎、压扁）与重度污染（污渍、残留液体）情况。数据集采用标准YOLO边界框格式（.txt文件），针对目标检测任务进行了精细化标注。其涵盖7个细粒度工业PET瓶分类：B1、B2.1、B2.2、B3、B4、B5及B6。本数据集可作为可靠且具有挑战性的基准测试集，用于开发循环经济倡议中的实时机器视觉与AI驱动的机器人分拣系统。