The dataset consists of images of the Kalo 1.5 heat cost allocator (HCA) and the Qundis HCA. The dataset has been created for the Reconcycle project. Find information at reconcycle.eu. The objects are positioned in different areas of the Reconcycle workcell, designed by JSI. The dataset has the following properties: 1577 images with resolution: 1450x1450 pixels (Basler camera ) 57 images with resolution: 848 x 480 pixels (Realsense D435 camera) The images have segmentation annotations labelled using the labelme software. The original labelme annotations are present and exported to COCO dataset format. The annotations are in the form of polygon segmentations. The included COCO train/test split is a 90/10 split. The images have been annotated with the following labels: hca_front hca_back hca_side1 hca_side2 battery pcb internals pcb_covered plastic_clip