Metric Learning With HORDE: High-Order Regularizer for Deep Embeddings
This work addresses robustness issues in visual search for applications like verification and zero-shot learning, though it is incremental as it builds on existing metric learning methods.
The paper tackled the problem of scattered deep features in metric learning for visual search tasks by introducing HORDE, a distribution-aware regularizer that enforces visually-close images to have similarly distributed features, resulting in state-of-the-art performance on four datasets.
Learning an effective similarity measure between image representations is key to the success of recent advances in visual search tasks (e.g. verification or zero-shot learning). Although the metric learning part is well addressed, this metric is usually computed over the average of the extracted deep features. This representation is then trained to be discriminative. However, these deep features tend to be scattered across the feature space. Consequently, the representations are not robust to outliers, object occlusions, background variations, etc. In this paper, we tackle this scattering problem with a distribution-aware regularization named HORDE. This regularizer enforces visually-close images to have deep features with the same distribution which are well localized in the feature space. We provide a theoretical analysis supporting this regularization effect. We also show the effectiveness of our approach by obtaining state-of-the-art results on 4 well-known datasets (Cub-200-2011, Cars-196, Stanford Online Products and Inshop Clothes Retrieval).