| dc.description.abstract | Dutch Landrace goats, primarily kept for their cultural and heritage value, have witnessed a decline in their population and a consequent loss of distinctive aesthetics. The Dutch Landrace goat traces its ancestry to a mere four individuals in 1958, complicating efforts to retrieve its original aesthetic features. In recent years, the Landelijke Fokkersclub Nederlandse Landgeiten has attempted to address these issues. The aim of the breeders club is to preserve the Dutch Landrace goat. Obtaining more specific knowledge on what morphological characteristics were dominant is crucial for the breeding program. For this study, a dataset of historic paintings and drawings containing goats was collected and provided by the breeding club. The objective was to discover, using Procrustes analysis, whether the 2D representations derived from the dataset could aid in enhancing the breeding program, using machine learning techniques and morphological analysis. The results indicate that, while Procrustes analysis provided useful insights into the morphological variations among Dutch Landrace goats, the reliability of the outcomes was impacted by variations in artistic representations and the quality of input images. Machine learning techniques were effective in extracting goats from images, yet were limited by the quality and realism of the data. These findings underscore the potential of integrating machine learning and geometric morphometrics in historical morphology research, while also highlighting the challenges associated with data noise and reproducibility. Further research in this domain is encouraged, as the possibilities to implement new machine leaning techniques become more accessible. | |
