Circle Loss: A Unified Perspective of Pair Similarity Optimization

TL;DR

This paper introduces Circle Loss, a flexible and unified loss function for deep feature learning that emphasizes poorly optimized similarities, improving performance across various recognition and retrieval tasks.

Contribution

The paper proposes Circle Loss, a novel similarity-based loss function that adaptively weights similarity scores, unifying class-level and pair-wise label learning approaches.

Findings

Circle Loss outperforms existing loss functions on multiple tasks.

It achieves state-of-the-art results in face recognition and image retrieval.

The method demonstrates improved convergence and flexibility.

Abstract

This paper provides a pair similarity optimization viewpoint on deep feature learning, aiming to maximize the within-class similarity $s_p$ and minimize the between-class similarity $s_n$. We find a majority of loss functions, including the triplet loss and the softmax plus cross-entropy loss, embed $s_n$ and $s_p$ into similarity pairs and seek to reduce $(s_n-s_p)$. Such an optimization manner is inflexible, because the penalty strength on every single similarity score is restricted to be equal. Our intuition is that if a similarity score deviates far from the optimum, it should be emphasized. To this end, we simply re-weight each similarity to highlight the less-optimized similarity scores. It results in a Circle loss, which is named due to its circular decision boundary. The Circle loss has a unified formula for two elemental deep feature learning approaches, i.e. learning with class-level labels and pair-wise labels. Analytically, we show that the Circle loss offers a more flexible optimization approach towards a more definite convergence target, compared with the loss functions optimizing $(s_n-s_p)$. Experimentally, we demonstrate the superiority of the Circle loss on a variety of deep feature learning tasks. On face recognition, person re-identification, as well as several fine-grained image retrieval datasets, the achieved performance is on par with the state of the art.