Delving Deep into Label Smoothing

Abstract

Label smoothing is an effective regularization tool for deep neural networks (DNNs), which generates soft labels by applying a weighted average between the uniform distribution and the hard label. It is often used to reduce the overfitting problem of training DNNs and further improve classification performance. In this paper, we aim to investigate how to generate more reliable soft labels. We present an Online Label Smoothing (OLS) strategy, which generates soft labels based on the statistics of the model prediction for the target category. The proposed OLS constructs a more reasonable probability distribution between the target categories and non-target categories to supervise DNNs. Experiments demonstrate that based on the same classification models, the proposed approach can effectively improve the classification performance on CIFAR-100, ImageNet, and fine-grained datasets. Additionally, the proposed method can significantly improve the robustness of DNN models to noisy labels compared to current label smoothing approaches.

arXiv comment12 pages, 7 figures, 12 tables

References in corpus (8)

• Res2Net: A New Multi-scale Backbone Architecture
• Classification with Noisy Labels by Importance Reweighting
• Object-Part Attention Model for Fine-grained Image Classification
• Low-resolution Face Recognition in the Wild via Selective Knowledge Distillation
• Weakly Supervised Fine-Grained Image Categorization
• ShakeDrop Regularization for Deep Residual Learning
• Real-Time Correlation Tracking via Joint Model Compression and Transfer
• A comparison of dense region detectors for image search and fine-grained classification

Cited by in corpus (15)

• Salient Objects in Clutter
• Convolutional Fine-Grained Classification with Self-Supervised Target Relation Regularization
• Exploration of Class Center for Fine-Grained Visual Classification
• Inverse Image Frequency for Long-tailed Image Recognition
• Event Transformer+. A multi-purpose solution for efficient event data processing
• AffectNet+: A Database for Enhancing Facial Expression Recognition with Soft-Labels
• Lesion-aware network for diabetic retinopathy diagnosis
• A Robust Initialization of Residual Blocks for Effective ResNet Training without Batch Normalization
• Improving Deep Representation Learning via Auxiliary Learnable Target Coding
• From CAD models to soft point cloud labels: An automatic annotation pipeline for cheaply supervised 3D semantic segmentation
• Enhancing Cross-Dataset Performance of Distracted Driving Detection With Score Softmax Classifier And Dynamic Gaussian Smoothing Supervision
• Semantic-embedded Similarity Prototype for Scene Recognition
• Robust and Efficient Adversarial Defense in SNNs via Image Purification and Joint Detection
• Adaptive Discriminative Regularization for Visual Classification
• Intra-class Patch Swap for Self-Distillation