Annotated datasets for waste electrical and electronic equipment identification and battery detection and classification 

XBAT+ Datasets for WEEE Identification and Battery Detection and Classification

Clickn on this Link to Download: zenodo.org/records/18022530 

(Email me if you face any issues: imrizvankhan@gmail.com)

This release of XBAT+ (Advanced Robotics and Artificial Intelligence for Critical Raw Materials Recycling in the Circular Economy) provides annotated training and test datasets composed of images of static WEEE devices with and/or without batteries. These training and test datasets are grouped as High-Quality (HQ) datasets, Varying-Quality (VQ) datasets and Red-Green-Blue (RGB) datasets. The HQ datasets consist of expert-selected X-ray images derived from the VQ datasets. The RGB datasets contain images acquired using an UltraSharp Dell webcam. The VQ datasets comprise X-ray images captured at multiple tube-voltage settings, without expert visual selection. 

As of  December 2025, XBAT+ defined 50 categories of battery-containing WEEE devices, manually sorted from cages of mixed WEEE, during visits at KMK Metals Recycling Limited and Mungret Civic Amenity Centre, as well as during Limerick City and University of Limerick collection events. This XBAT+ 1.0 release was extracted from 15 categories (classes), each represented by more than five devices. Within each of these 15 categories, 20% of images were selected for the test subset. The overall test set was then formed by combining these per-category subsets, to ensure the test data are representative across categories. The remaining images from each category were combined to form the training set.

The uploaded ZIP files, raw_XBAT+_v1.0_ ... .zip and res_XBAT+_v1.0_ ... .zip, include raw and resized data organized as follows:

• HQ Datasets

  • Test data: 91 images, 91 labels

  • Training data: 330 images, 330 labels

• RGB Datasets

  • Test data: 91 images, 91 labels

  • Training data: 330 images, 330 labels

• VQ Datasets

  • Test data: 482 images, 482 labels

  • Training data: 1,721 images, 1,721 labels

These datasets are fully described in the accompanying data descriptor, titled “Annotated datasets for waste electrical and electronic equipment identification and battery detection and classification”. Users of the datasets are encouraged to cite the associated data descriptor publication. Note that these preliminary XBAT+ datasets are intended to support research and development in battery-containing WEEE identification, battery presence detection, and battery chemistry classification.