Exponential Discriminative Metric Embedding in Deep Learning

2018/03/07 by Bowen Wu, Zhangling Chen, Wu, Bowen +5
Computer Science · Mathematics · #Artificial intelligence #Class (philosophy) #Compact space #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Convolutional neural network #Deep learning #Discriminative model #Embedding #FOS: Computer and information sciences #Face and Expression Recognition #Face recognition and analysis #Feature learning #Feature vector #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine learning #Margin (machine learning) #Mathematics #Metric (unit) #Object (grammar) #Pattern recognition (psychology) #Video Surveillance and Tracking Methods #cs.CV #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.1803.02504

arxiv created 2018/03/07 · openalex publication_date 2018/03/07 · arxiv updated 2018/03/08 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

Abstract

With the remarkable success achieved by the Convolutional Neural Networks (CNNs) in object recognition recently, deep learning is being widely used in the computer vision community. Deep Metric Learning (DML), integrating deep learning with conventional metric learning, has set new records in many fields, especially in classification task. In this paper, we propose a replicable DML method, called Include and Exclude (IE) loss, to force the distance between a sample and its designated class center away from the mean distance of this sample to other class centers with a large margin in the exponential feature projection space. With the supervision of IE loss, we can train CNNs to enhance the intra-class compactness and inter-class separability, leading to great improvements on several public datasets ranging from object recognition to face verification. We conduct a comparative study of our algorithm with several typical DML methods on three kinds of networks with different capacity. Extensive experiments on three object recognition datasets and two face recognition datasets demonstrate that IE loss is always superior to other mainstream DML methods and approach the state-of-the-art results.

Citations

• Learning Face Representation from Scratch
• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Object Detectors Emerge in Deep Scene CNNs
• Improving neural networks by preventing co-adaptation of feature detectors
• High-Performance Neural Networks for Visual Object Classification
• Gradient-based learning applied to document recognition
• DeepFace: Closing the Gap to Human-Level Performance in Face Verification
• Wide Residual Networks
• Large-Margin Softmax Loss for Convolutional Neural Networks
• Deep Residual Learning for Image Recognition
• Generalizing Pooling Functions in Convolutional Neural Networks: Mixed, Gated, and Tree

Related

• A Framework to Enhance Generalization of Deep Metric Learning methods using General Discriminative Feature Learning and Class Adversarial Neural Networks
• Feature Representation in Deep Metric Embeddings
• Scalable Angular Discriminative Deep Metric Learning for Face Recognition
• Anchor-based Nearest Class Mean Loss for Convolutional Neural Networks
• On Learning Density Aware Embeddings