DiVA: Diverse Visual Feature Aggregation for Deep Metric Learning
This addresses the generalization issue in DML for computer vision applications, offering a novel approach to improve performance on unknown classes, though it is incremental in building upon existing DML frameworks.
The paper tackles the problem of deep metric learning (DML) models specializing too much on training classes, which limits generalization to unknown test classes, by proposing DiVA, which aggregates diverse visual features through multiple complementary learning tasks, achieving state-of-the-art performance on established DML benchmarks.
Visual Similarity plays an important role in many computer vision applications. Deep metric learning (DML) is a powerful framework for learning such similarities which not only generalize from training data to identically distributed test distributions, but in particular also translate to unknown test classes. However, its prevailing learning paradigm is class-discriminative supervised training, which typically results in representations specialized in separating training classes. For effective generalization, however, such an image representation needs to capture a diverse range of data characteristics. To this end, we propose and study multiple complementary learning tasks, targeting conceptually different data relationships by only resorting to the available training samples and labels of a standard DML setting. Through simultaneous optimization of our tasks we learn a single model to aggregate their training signals, resulting in strong generalization and state-of-the-art performance on multiple established DML benchmark datasets.