Although computer vision is being developed across several fields, no wellestablished dataset currently exists for visual insulated pipeline inspection. The lack of such a dataset negatively affects research progress, but most importantly limits numerous real-world applications, especially industrial ones. To address this, we introduce, to the best of our knowledge, the first publicly available comprehensive Visual Insulated Pipeline (VIP) dataset, comprising 3,400 images across seven European locations and annotated for three computer vision tasks: (i) Pipeline region semantic segmentation, (ii)Pipeline damage semantic segmentation, and (iii) Pipeline damage object detection. Moreover, we implement state-of-the-art algorithms for these tasks on VIP and empirically demonstrate, both quantitatively and qualitatively, that there is ample room for improvement. Therefore, the proposed dataset constitutes a valuable research benchmark for the development of novel computer vision algorithms, particularly in the domain of visual insulated pipeline inspection. In addition, knowledge gained from this domain may be transferable to other domains (e.g., industrial ones), although this was not tested in this paper.