Circle Loss: A Unified Perspective of Pair Similarity Optimization

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.

References in corpus (16)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Deep Residual Learning for Image Recognition
• In Defense of the Triplet Loss for Person Re-Identification
• NormFace: L2 Hypersphere Embedding for Face Verification
• Large-Margin Softmax Loss for Convolutional Neural Networks
• SphereFace: Deep Hypersphere Embedding for Face Recognition
• L2-constrained Softmax Loss for Discriminative Face Verification
• CosFace: Large Margin Cosine Loss for Deep Face Recognition
• Learning Deep Embeddings with Histogram Loss
• MS-Celeb-1M: A Dataset and Benchmark for Large-Scale Face Recognition
• Multi-Similarity Loss with General Pair Weighting for Deep Metric Learning
• Heated-Up Softmax Embedding
• AdaCos: Adaptively Scaling Cosine Logits for Effectively Learning Deep Face Representations
• Attention-based Ensemble for Deep Metric Learning
• Deep Metric Learning via Facility Location
• Softmax Dissection: Towards Understanding Intra- and Inter-class Objective for Embedding Learning

Cited by in corpus (19)

• FastReID: A Pytorch Toolbox for General Instance Re-identification
• ERNIE-ViL: Knowledge Enhanced Vision-Language Representations Through Scene Graph
• Parameter-Efficient Person Re-identification in the 3D Space
• A Contrastive Variational Graph Auto-Encoder for Node Clustering
• FaceX-Zoo: A PyTorch Toolbox for Face Recognition
• Dual-Refinement: Joint Label and Feature Refinement for Unsupervised Domain Adaptive Person Re-Identification
• Attribute Adaptive Margin Softmax Loss using Privileged Information
• Prototype Memory for Large-scale Face Representation Learning
• Ranked List Loss for Deep Metric Learning
• EqCo: Equivalent Rules for Self-supervised Contrastive Learning
• Crossover Learning for Fast Online Video Instance Segmentation
• EQFace: A Simple Explicit Quality Network for Face Recognition
• Attribute-guided Feature Extraction and Augmentation Robust Learning for Vehicle Re-identification
• Combating Ambiguity for Hash-code Learning in Medical Instance Retrieval
• Multi-Facet Recommender Networks with Spherical Optimization
• Information Theory-Guided Heuristic Progressive Multi-View Coding
• Dual-Tuning: Joint Prototype Transfer and Structure Regularization for Compatible Feature Learning
• Dynamic Metric Learning: Towards a Scalable Metric Space to Accommodate Multiple Semantic Scales
• Boosting Unconstrained Face Recognition with Auxiliary Unlabeled Data