The increasing demand for effective and safe methods of inspecting and monitoring industrial buildings has led to the adoption of advanced technologies. This article proposes an innovative methodology that combines advanced computer vision techniques and unmanned aerial vehicles (UAVs) for the condition assessment of large-scale industrial buildings, addressing typical challenges associated with data acquisition and automatic identification of pathologies. The methodology comprises the following stages: (i) data acquisition and preprocessing, (ii) data labelling, (iii) training a Deep Convolutional Neural Network, and (iv) creation of a 3D photogrammetric As-Is model. In the first stage, a georeferenced photographic survey is carried out with the UAV, capturing high-resolution and close-up images of the entire building envelope. These images are then processed using Structure from Motion and Multi-View Stereo algorithms to obtain a georeferenced photogrammetric model. In the data labelling stage, several images are previously labelled. For that, in the next stage, be able to automatically identify the various pathologies in supervised learning. Therefore, the labelled images are used to train and validate a deep learning model specifically designed for damage detection and segmentation. This model is optimized to recognize and classify different types of damage, such as corrosion, mechanical damage, and water accumulation. In the last stage, the images collected by the UAV are tested on the optimized network, allowing for the inference of pathologies. Additionally, 3D projections of the pathologies over the photogrammetric model are performed using a dedicated Ray Casting technique, which enables an improved visualization and analysis of the affected areas. The proposed methodology has been successfully tested and validated in a real industrial building, showing its significant contribution to the application of preventive maintenance strategies and promoting the efficient management of specific types of buildings. By integrating these advanced technologies and methodologies, the study aims to provide a comprehensive and efficient solution for industrial building inspection and maintenance. The resulting digital model and damage identification methodology will not only improve the accuracy and reliability of inspections but also contribute to the overall safety and longevity of industrial infrastructure.