Quadruplet Network with One-Shot Learning for Visual Tracking
As a discriminative method of one-shot learning, Siamese deep network allows recognizing an object from a single exemplar with the same class label. However, it does not take the advantage of the underlying structure and relationship among a multitude of instances since it only relies on pairs of instances for training. In this paper, we propose a quadruplet deep network to examine the potential connections among the training instances, aiming to achieve a more powerful representation. We design four shared networks that receive multi-tuple of instances as inputs and are connected by a novel loss function consisting of pair-loss and triplet-loss. According to the similarity metric, we select the most similar and the most dissimilar instances as the positive and negative inputs of triplet loss from each multi-tuple. We show that this scheme improves the training performance and convergence speed. Furthermore, we introduce a new weighted pair loss for an additional acceleration of the convergence. We demonstrate promising results for model-free tracking-by-detection of objects from a single initial exemplar in the Visual Object Tracking benchmark.
READ FULL TEXT