Deep learning approach to detect and visualize sexual dimorphism in monomorphic species

Abstract

Sex recognition is facilitated by dimorphism in some traits. However, humans often fail to find the traits that allow distinction between sexes in other species. Deep learning has shown great application potential in identifying cryptic differences between sexes, but it has rarely been used to assess sexual dimorphism. In this study, the ability of a fine-tuned classification neural network, which is known as EfficientNet, to find differences between sexes in a species that appears monomorphic to humans, such as the sociable weaver, Philetairus socius, was evaluated. In addition, the benefits of the Grad-CAM visualization technique were assessed to understand which parts of the head of the individuals are used by the network to differentiate the sexes. We trained 10-fold cross-validation models on more than 4500 pictures of the head from more than 1300 individuals. Results show that the network can predict the sex of sociable weavers with an accuracy of 76%, which is considerably higher than humans' performance (56%). Moreover, the model was similarly good at predicting females and males. When interpreting the probability of being classified to one sex, our results further reveal the effect of the interaction of sex with age on the confidence score of the models, which shows that younger males are less masculine than older ones, and older females are more masculine than younger ones. Finally, using Grad-CAM, we found that the model mostly used the bill region to predict the sex of individuals. Overall, this work shows the potential application of artificial intelligence as a noninvasive sexing tool, surpassing human capabilities and aiding in pinpointing potential cryptic dimorphic body parts that have yet to be identified. Half of the world's bird species appear sexually monomorphic to humans, and re-evaluation of species dimorphism with this type of methods could deepen our understanding of topics such as sex-specific selection on animal traits, behavioural differences and sex ratio variation across time.