Deep Label Distribution Learning with Label Ambiguity

Abstract

Convolutional Neural Networks (ConvNets) have achieved excellent recognition performance in various visual recognition tasks. A large labeled training set is one of the most important factors for its success. However, it is difficult to collect sufficient training images with precise labels in some domains such as apparent age estimation, head pose estimation, multi-label classification and semantic segmentation. Fortunately, there is ambiguous information among labels, which makes these tasks different from traditional classification. Based on this observation, we convert the label of each image into a discrete label distribution, and learn the label distribution by minimizing a Kullback-Leibler divergence between the predicted and ground-truth label distributions using deep ConvNets. The proposed DLDL (Deep Label Distribution Learning) method effectively utilizes the label ambiguity in both feature learning and classifier learning, which help prevent the network from over-fitting even when the training set is small. Experimental results show that the proposed approach produces significantly better results than state-of-the-art methods for age estimation and head pose estimation. At the same time, it also improves recognition performance for multi-label classification and semantic segmentation tasks.

Accepted by IEEE TIP 2017. Projects page, see http://lamda.nju.edu.cn/gaobb/projects/DLDL.html

References in corpus (4)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Efficient Inference in Fully Connected CRFs with Gaussian Edge Potentials
• Fully Convolutional Networks for Semantic Segmentation
• Instance-Aware Hashing for Multi-Label Image Retrieval

Cited by in corpus (59)

• Label Distribution Learning Forests
• Probabilistic End-to-end Noise Correction for Learning with Noisy Labels
• An End-to-End Deep Learning Histochemical Scoring System for Breast Cancer Tissue Microarray
• Co-Correcting: Noise-tolerant Medical Image Classification via mutual Label Correction
• EEG-Based Emotion Recognition Using Regularized Graph Neural Networks
• Deep Weakly-Supervised Learning Methods for Classification and Localization in Histology Images: A Survey
• Deep Ordinal Regression with Label Diversity
• Learning in an Uncertain World: Representing Ambiguity Through Multiple Hypotheses
• Seeking Subjectivity in Visual Emotion Distribution Learning
• CASTLE: Regularization via Auxiliary Causal Graph Discovery
• DPGN: Distribution Propagation Graph Network for Few-shot Learning
• Navigation-Oriented Scene Understanding for Robotic Autonomy: Learning to Segment Driveability in Egocentric Images
• MetaAge: Meta-Learning Personalized Age Estimators
• Fine-Grained Age Estimation in the wild with Attention LSTM Networks
• Beyond Cats and Dogs: Semi-supervised Classification of fuzzy labels with overclustering
• Fuzzy Overclustering: Semi-Supervised Classification of Fuzzy Labels with Overclustering and Inverse Cross-Entropy
• Quantifying Facial Age by Posterior of Age Comparisons
• PhysFormer: Facial Video-based Physiological Measurement with Temporal Difference Transformer
• End-to-End Label Uncertainty Modeling in Speech Emotion Recognition using Bayesian Neural Networks and Label Distribution Learning
• C3AE: Exploring the Limits of Compact Model for Age Estimation
• Neural-network-powered pulse reconstruction from one-dimensional interferometric cross-correlation traces
• Generalized Label Enhancement with Sample Correlations
• Uncertainty-aware Score Distribution Learning for Action Quality Assessment
• Soft-ranking Label Encoding for Robust Facial Age Estimation
• Learning to Infer Unseen Single-/Multi-Attribute-Object Compositions with Graph Networks
• From CAD models to soft point cloud labels: An automatic annotation pipeline for cheaply supervised 3D semantic segmentation
• Learning Expectation of Label Distribution for Facial Age and Attractiveness Estimation
• A Coupled Evolutionary Network for Age Estimation
• Tweets Sentiment Analysis via Word Embeddings and Machine Learning Techniques
• Repetitive Reprediction Deep Decipher for Semi-Supervised Learning
• Distilling Virtual Examples for Long-tailed Recognition
• Non-parametric Uni-modality Constraints for Deep Ordinal Classification
• A Circular-Structured Representation for Visual Emotion Distribution Learning
• Facial Pose Estimation by Deep Learning from Label Distributions
• Deep Regression Forests for Age Estimation
• Bounding-box deep calibration for high performance face detection
• Deep Differentiable Random Forests for Age Estimation
• Many-to-One Distribution Learning and K-Nearest Neighbor Smoothing for Thoracic Disease Identification
• Sample Efficient Learning of Image-Based Diagnostic Classifiers Using Probabilistic Labels
• On Modelling Label Uncertainty in Deep Neural Networks: Automatic Estimation of Intra-observer Variability in 2D Echocardiography Quality Assessment
• Evaluating AI systems under uncertain ground truth: a case study in dermatology
• Learning Probabilistic Ordinal Embeddings for Uncertainty-Aware Regression
• Computational Emotion Analysis From Images: Recent Advances and Future Directions
• Facial Emotion Recognition with Noisy Multi-task Annotations
• Neighborhood-based Pooling for Population-level Label Distribution Learning
• Medical Image Segmentation with Limited Supervision: A Review of Deep Network Models
• Learning to Rectify for Robust Learning with Noisy Labels
• PML: Progressive Margin Loss for Long-tailed Age Classification
• Improving Label Quality by Jointly Modeling Items and Annotators
• Deep Ordinal Regression using Optimal Transport Loss and Unimodal Output Probabilities
• Learning and Exploiting Interclass Visual Correlations for Medical Image Classification
• Deep Ordinal Regression for Pledge Specificity Prediction
• Amalgamating Filtered Knowledge: Learning Task-customized Student from Multi-task Teachers
• Improved Mix-up with KL-Entropy for Learning From Noisy Labels
• Self-Paced Deep Regression Forests for Facial Age Estimation
• Computationally Efficient Wasserstein Loss for Structured Labels
• Representing and Detecting Label Ambiguity in IMU-Based Exercise Evaluation
• Diagnostic Uncertainty Calibration: Towards Reliable Machine Predictions in Medical Domain
• A Framework of Transferring Structures Across Large-scale Information Networks