Boosting Standard Classification Architectures Through a Ranking Regularizer
This work addresses the challenge of boosting standard classification architectures for fine-grained and imbalanced recognition tasks, though it is incremental as it builds on existing methods like triplet loss.
The paper tackles the problem of improving classification performance by using triplet loss as a feature embedding regularizer, resulting in steady improvements on five fine-grained recognition datasets and significant gains on an imbalanced video dataset.
We employ triplet loss as a feature embedding regularizer to boost classification performance. Standard architectures, like ResNet and Inception, are extended to support both losses with minimal hyper-parameter tuning. This promotes generality while fine-tuning pretrained networks. Triplet loss is a powerful surrogate for recently proposed embedding regularizers. Yet, it is avoided due to large batch-size requirement and high computational cost. Through our experiments, we re-assess these assumptions. During inference, our network supports both classification and embedding tasks without any computational overhead. Quantitative evaluation highlights a steady improvement on five fine-grained recognition datasets. Further evaluation on an imbalanced video dataset achieves significant improvement. Triplet loss brings feature embedding characteristics like nearest neighbor to classification models. Code available at \url{http://bit.ly/2LNYEqL}.