Relevance. Core sample analysis is fundamental for studying oil and gas reservoirs. However, conventional core investigation methods – especially UV fluorescence-based oil saturation assessment – are highly labor-intensive, human-dependent, and has no clear guidelines. Increasing volumes of field data exacerbate this issue, necessitating automated, objective, and scalable solutions. Aim. To develop a hybrid algorithm for determining core oil saturation using macro-photographs of rocks in daylight and ultraviolet light, along with its software implementation. Methods. Image processing, computer vision, and deep learning methods. For UV fluorescence classification, we proposed a multimodal model based on ResNet-34. It processes six channels simultaneously: RGB daylight + RGB UV light. The software was implemented as an intelligent information system using a microservice architecture. Results. We developed a UV fluorescence classification algorithm achieving 90% accuracy in identifying fluorescence type and 80% accuracy in intensity assessment. The implemented intelligent information system processes high-resolution images (4000×4000 px) and reduces storage requirements by 20–40% by retaining only essential segmented image parts. Conclusions. The combined approach – integrating multi-image analysis and flexible microservices – appears promising. It not only automates and objectifies oil saturation assessment but also establishes a foundation for future digital petrophysics research. This solution enables more accurate saturation determination, faster data analysis, and optimized storage, which could benefit the oil and gas industry.