Abstract:This work aims to address the image classification problem under open-set protocol: classes in test set do not appear in the training set. Intuitively, convolutional Neural Network (CNN) with softmax loss is a straight-forward solution. However, the unknown class (is not predefined in the training set) makes the boundaries of intra-class and inter-class more blurred, which brings more challenges for image classification. Although some softmax variants, such as center loss, CosFace loss etc., focus on learning discriminative features by minimizing the intra-class distance, they do not explicitly maximize inter-class distance, which is more important for open-set problem verified by our experiments. Besides, even though deep metric learning, such as with the contrastive loss and the triplet loss, can learn discriminative features of intra-class and inter-class, it needs a time-consuming image sampling process during training. In this paper, we propose a novel normalized maximal margin (NMM) loss for open-set image classification, which not only explicitly minimizes intra-class distance and maximizes inter-class distance, but also defines their margins. Specially, after analyzing the advantage of angular space that the softmax loss normalized by the feature and weights through geometric interpretation, we make NMM work in angular space. Then, the validity of NMM for discriminative features learning is demonstrated from the view of geometric interpretation as well. After that, we innovatively determine the upper bound of inter-class margin by theoretical analysis. Finally, extensive experiments are conducted on popular datasets: CIFAR-100 (object recognition), ImageNet (image classification), LFW (face recognition) and MSMT17 (person re-identification) to verify the effectiveness of NMM. The experimental results show that NMM achieves very competitive performance.

Class Anchor Clustering: a Loss for Distance-based Open Set Recognition

Human Activity Recognition Machine With an Anchor-Based Loss Function

Wasserstein Distance-based Expansion of Low-Density Latent Regions for Unknown Class Detection

Learning from Noisy Anchors for One-Stage Object Detection.

Cascading Unknown Detection with Known Classification for Open Set Recognition

Self-paced Deep Clustering with Learning Loss

Class Anchor Margin Loss for Content-Based Image Retrieval

Boosting Deep Open World Recognition by Clustering

Orientational Clustering Learning for Open-Set Hyperspectral Image Classification

Cluster-based Adversarial Decision Boundary for domain-adaptive open set recognition

From anomaly detection to open set recognition: Bridging the gap

Normalized Maximal Margin Loss for Open-Set Image Classification

Open Set Intrusion Event Recognition Using Anchor Point Learning for Distributed Optical Fiber System

Deep Clustering with Incomplete Noisy Pairwise Annotations: A Geometric Regularization Approach

LapNet : Automatic Balanced Loss and Optimal Assignment for Real-Time Dense Object Detection

Class-Specific Semantic Reconstruction for Open Set Recognition

LAG: Layered Objects to Generate Better Anchors for Object Detection in Aerial Images

P-ODN: Prototype based Open Deep Network for Open Set Recognition

Learning to Match Anchors for Visual Object Detection

Open-Set Face Recognition with Maximal Entropy and Objectosphere Loss

FreeAnchor: Learning to Match Anchors for Visual Object Detection