Background: Sperm morphology assessment is central to bull breeding soundness and fertility evaluation; however, current spermiogram assessment are labour-intensive, subjective, and prone to inter- and intra-observer variation. Progress in artificial intelligence (AI)-assisted sperm phenotyping has been constrained by the lack of large, expertly annotated datasets of high-resolution images acquired using gold-standard differential interference contrast (DIC) microscopy, with detailed morphological annotations. To address this gap, this study describes the development of an expanding sperm morphology dataset and deep-learning models for automated sperm analysis, establishing the foundation for AI-driven morphology classification and prediction of bull fertility. Methods: Individual bull spermatozoa were imaged with a 100x oil immersion objective lens using DIC microscopy. An annotation framework was established to label sperm anatomical structures (head, midpiece, tail and whole sperm) and 32 morphological categories under expert supervision. Annotated images were divided into training (70%), validation (15%) and test (15%) sets. Multiple deep-learning segmentation architectures, including Mask2Former, EoMT-DINO, and LoRA-EoMT-DINO variants, were trained and evaluated using intersection over union (IoU). Results: The dataset currently comprises 638 annotated sperm images and continues to expand in size and morphological diversity. At the time of analysis, 490 annotated images were available for model development and evaluation. Across the evaluated architectures, the Mask2Former achieved the highest overall performance, with segmentation IoU scores of up to 0.94 for sperm heads, 0.86 for midpieces, 0.82 for tails, and 0.90 for whole-sperm regions. Automated segmentation enabled reliable identification of sperm structures across diverse morphological presentations. Conclusion: We present a high-resolution sperm morphology resource that combines detailed structural and morphological annotation with AI-driven image analysis. Ongoing expansion of the dataset will support increasingly accurate and comprehensive characterisation of sperm phenotypes, improve understanding of morphological features associated with reproductive performance, and enable model development for morphology classification, bull fertility prediction, and breeding outcome assessment.