IPD Dataset

The dataset comprises 357 images of insulated pipelines captured from industrial and public facilities in Europe. Each image was pixel-level annotated internally with binary masks, taking into account three types of pipeline insulation damage: i) hole, ii) open insulation, and iii) dent. The damage types all have the same pixel-level label (i.e., 1), and there is no distinction in the masks between different types of damage. To advance research in the relevant field, AUTH made the dataset publicly available for research purposes via the AIIA lab. All requests for access to/use of the dataset must be submitted in writing by researchers. To receive a copy of the dataset, the researcher (hereinafter, the Signatory) must agree to the conditions below and sign this license agreement. Subsequently, email it to Prof. Ioannis Pitas (using “SIMAR – IPD dataset availability” as e-mail subject) so as to receive further instructions for downloading.