Computer-Aided Sperm Analysis (CASA) systems play a critical role in diagnosing fertility issues. They are essential for treating infertility in humans and are equally valuable in the domain of domestic animal reproduction, where they support studies on sperm competition and wildlife semen assessment. However, the lack of publicly available datasets and the difficulties associated with data labelling constitute a significant challenge. In this paper, we propose a solution to address this challenge by introducing a synthetic data generation approach based on a parametric model. We demonstrate the effectiveness of our approach by successfully addressing two key sperm analysis tasks: sperm count and sperm motility assessment. Our proposed system uses synthetic data for training and is validated with real-world boar sperm images. The results obtained show a mean average precision (mAP) of 84.3% for the sperm count task, and a multiple object tracking accuracy (MOTA) of 80.5% for motility estimation task. Furthermore, the computational requirements of our solution are minimal, enabling its execution on embedded systems and facilitating its integration into professional CASA systems.
