Self-Supervised Approach to Addressing Zero-Shot Learning Problem
This addresses a domain-specific challenge in entomology for species identification, but it is incremental as it applies existing methods to a new dataset.
The paper tackled the problem of zero-shot learning for nearly indistinguishable bumblebee species using a self-supervised Siamese neural network with contrastive loss, achieving a 61% F1-score on zero-shot instances and an 11% improvement on classes overlapping with the training set.
In recent years, self-supervised learning has had significant success in applications involving computer vision and natural language processing. The type of pretext task is important to this boost in performance. One common pretext task is the measure of similarity and dissimilarity between pairs of images. In this scenario, the two images that make up the negative pair are visibly different to humans. However, in entomology, species are nearly indistinguishable and thus hard to differentiate. In this study, we explored the performance of a Siamese neural network using contrastive loss by learning to push apart embeddings of bumblebee species pair that are dissimilar, and pull together similar embeddings. Our experimental results show a 61% F1-score on zero-shot instances, a performance showing 11% improvement on samples of classes that share intersections with the training set.