Abstract: The Elephant Identification project aims to automate the identification of individual elephants using machine learning techniques applied to ground-based images and videos. By replacing traditional, labour-intensive and time-consuming methods like Rapid Elephant Population Assessments (REPA), the project seeks to support conservation efforts through a more efficient, automated system. The process involves pre-processing images, extracting features, and applying heuristic and hierarchical clustering to distinguish individual elephants. Amongst the evaluated techniques, applying ResNet-50 on Local Binary Pattern (LBP) images proved to be more effective in uniquely identifying elephants as it emphasises texture and pattern information. Upon visual evaluation, approximately 50% of the observed images are correctly classified. However, the lack of a labelled dataset limits the evaluation and validation of the model. External dataset is helpful but introduces subjectivity and inconsistency. To address this, a collaborative application has been developed to incorporate user feedback on clustering outcomes, aligning with supervised and semi-supervised learning principles. This feedback mechanism enables the model to evolve through continuous user interaction, contributing to data labelling and iterative improvement.