Advanced Computer Vision Techniques for Reliable Gender Determination in Budgerigars (Melopsittacus Undulatus)

Determining the gender of budgerigars by observing cere color can be unreliable and subjective, especially in young birds or those with atypical color variations. In this study, a comprehensive approach to Budgerigar (Melopsittacus undulatus) Gender Recognition was presented by applying advanced techniques of computer vision. Diverse methodologies were added to set up a reliable system to identify the gender of budgerigars, focusing on images of yellow, white, green, and blue budgerigars based on cere color. A dataset was prepared, and multiple deep learning models, including YOLOvge, YOLOv8x, YOLOv8m, YOLOv5x, YOLOv5m, VGG19, ResNet152, EfficientNet, and AlexNet, were experimented on, focusing on object detection to locate the cere and determine gender accurately. The YOLOvge state-of-the-art model showed the best performance. The results indicate wide implications for avian research, providing a new tool for gender identification for budgerigars and possibly other parakeet species.