Photography of drill core is one of the most long-lived and widely available sources of geological information, having been routinely collected in exploration and mining for decades. Repositories of core imagery contain vast amounts of geological information that are often underutilised, particularly for quantitative analysis. Barriers to using drill core photography as a quantitative data source include inconsistency in image quality, time-consuming manual processing of images, and limitations of traditional image-analysis methods to deal with geological variability. The recent advent of readily accessible high-powered computing coupled with significant advances in computer vision technologies are breaking these barriers and enabling consistent, high-quality quantitative analysis of core photography. This paper presents several examples of the application deep learning applied to core photography, showing that certain geological workflows can be augmented with data driven models. Image segmentation will be demonstrated by cropping and de-skewing rows from core-boxes, then further segmenting coherent rock. The results of this will be used to calculate Rock Quality Designation (RQD). Unsupervised image similarity geological domaining will also be discussed. The validity of these models will be shown to approach and surpass human-level accuracy and fidelity. CITATION: Webb, L, Grujic, M and Truong, D T, 2019. The application of deep learning to extract geological information from drill core imagery, in Proceedings Mining Geology 2019, pp 435442 (The Australasian Institute of Mining and Metallurgy: Melbourne).