Alam, Md. AshrafulTanvir, Md. Mehedi HasanAbir, Md. Ashiqur RahmanPrioty, Tasfia TasnimGomes, Arnab AnthonyFariha, Sadia Tasnim2026-01-212026-01-2120252025-10ID 22101107ID 22101650ID 23241101ID 22101351ID 21301294http://hdl.handle.net/10361/27473Cataloged from PDF version of thesis.Includes bibliographical references (pages 43-46).This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2025.Image geolocalization, the task of determining the geographic origin of a given image, remains a formidable challenge due to the immense variability of global landscapes and the subtle visual cues that indicate specific locations. This research aims to introduce a novel approach that seeks to surpass the accuracy of current state-ofthe- art models. By leveraging an highly diverse dataset, integrating cutting-edge vision transformer architectures, optimizing the training process, and systematic review and fine-tuning, this approach achieves significantly improved performance in image geolocalization. Our model demonstrates an exceptional capacity to generalize to previously unseen locations, even under complex conditions such as varied lighting, diverse weather patterns, and other environmental challenges, marking an advancement over prior methodologies in this domain.55 pagesenBRAC University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission.Image geolocalizationImage classificationComputer visionVision transformersContrastive learningStreet view analysisVisual place recognitionSpatial data analysisGeographic coordinate predictionMachine learningGeographic information systems.Spatial analysis (Statistics).Geospatial data--Computer processing.Global Positioning System--Data processing.Image processing.Photogrammetry--Digital techniques.Thesis