Digital cultural heritage repositories require image classification and retrieval techniques. Research communities currently shift their focus from handcrafted features towards unsupervised feature learning, with the aim to apply these techniques to both image and video analysis. The main idea of this approach consists in using large amounts of unlabelled images to learn image representations. This step is usually accomplished using deep neural architectures. Since unsupervised feature learning is based on availability of large image datasets and computationally complex algorithms, it has become commonplace to use HPC systems for their implementation. Photogrammetry is quite an old discipline and even its digital version has a long tradition of studies and applications. The specific topic of photogrammetric applications to landscape studies constitutes probably one of the most important and promising research branch, i.e. giving archaeologists the possibility to get a digital measurable model of the object of their studies. Recent developments in the Computer Vision community are presently offering new automated procedures for both image orientation and 3D reconstruction purposes at different scales. The advent of drones for low altitude aerial photography makes it even easier to collect endless number of overlapping photographs for different purposes and final ground resolution of the 3D model. A specific software for automated image georeferencing, developed at IMS/FORTH, will be integrated into VI-SEEM to allow for large image-set to be automatically matched with a given orthophoto and geo-referenced in a completely automated way, without human input and completely controllable output.