Training Convolutional Networks with Noisy Labels
arXiv:1406.2080
Abstract
The availability of large labeled datasets has allowed Convolutional Network models to achieve impressive recognition results. However, in many settings manual annotation of the data is impractical; instead our data has noisy labels, i.e. there is some freely available label for each image which may or may not be accurate. In this paper, we explore the performance of discriminatively-trained Convnets when trained on such noisy data. We introduce an extra noise layer into the network which adapts the network outputs to match the noisy label distribution. The parameters of this noise layer can be estimated as part of the training process and involve simple modifications to current training infrastructures for deep networks. We demonstrate the approaches on several datasets, including large scale experiments on the ImageNet classification benchmark.
Accepted as a workshop contribution at ICLR 2015
References in corpus (2)
Cited by in corpus (162)
- Temporal Ensembling for Semi-Supervised Learning
- Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labels
- Training Deep Neural Networks on Noisy Labels with Bootstrapping
- MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labels
- Deep Learning is Robust to Massive Label Noise
- Residual Attention Network for Image Classification
- Dimensionality-Driven Learning with Noisy Labels
- Masking: A New Perspective of Noisy Supervision
- A Light CNN for Deep Face Representation with Noisy Labels
- Robust Loss Functions under Label Noise for Deep Neural Networks
- Symmetric Cross Entropy for Robust Learning with Noisy Labels
- Decoupling "when to update" from "how to update"
- Supervised Contrastive Learning
- Making Deep Neural Networks Robust to Label Noise: a Loss Correction Approach
- Curriculum Loss: Robust Learning and Generalization against Label Corruption
- Learning From Noisy Singly-labeled Data
- Learning with Noise: Enhance Distantly Supervised Relation Extraction with Dynamic Transition Matrix
- Facial Expression Recognition from World Wild Web
- Breaking the Dilemma of Medical Image-to-image Translation
- Probabilistic End-to-end Noise Correction for Learning with Noisy Labels
- Webly Supervised Learning of Convolutional Networks
- FoodX-251: A Dataset for Fine-grained Food Classification
- Learning from Noisy Labels with Distillation
- Unsupervised Learning with Stein's Unbiased Risk Estimator
- Low-Quality Training Data Only? A Robust Framework for Detecting Encrypted Malicious Network Traffic
- Learning From Noisy Large-Scale Datasets With Minimal Supervision
- On the Effects of Different Types of Label Noise in Multi-Label Remote Sensing Image Classification
- Denoising Multi-Source Weak Supervision for Neural Text Classification
- EEG-Based Emotion Recognition Using Regularized Graph Neural Networks
- Weakly Supervised Vessel Segmentation in X-ray Angiograms by Self-Paced Learning from Noisy Labels with Suggestive Annotation
- Deep Weakly-Supervised Learning Methods for Classification and Localization in Histology Images: A Survey
- An Entropic Optimal Transport Loss for Learning Deep Neural Networks under Label Noise in Remote Sensing Images
- Loss factorization, weakly supervised learning and label noise robustness
- Label Noise Types and Their Effects on Deep Learning
- BadLabel: A Robust Perspective on Evaluating and Enhancing Label-noise Learning
- CleanNet: Transfer Learning for Scalable Image Classifier Training with Label Noise
- Learning from Noisy Labels for Entity-Centric Information Extraction
- L_DMI: An Information-theoretic Noise-robust Loss Function
- Confidence Regularized Self-Training
- Products-10K: A Large-scale Product Recognition Dataset
- Iterative Learning with Open-set Noisy Labels
- Learning From Noisy Labels By Regularized Estimation Of Annotator Confusion
- Avoiding Your Teacher's Mistakes: Training Neural Networks with Controlled Weak Supervision
- Transfer Learning with Label Noise
- What Do Neural Networks Learn When Trained With Random Labels?
- Uncertainty Based Detection and Relabeling of Noisy Image Labels
- Auxiliary Image Regularization for Deep CNNs with Noisy Labels
- The Unreasonable Effectiveness of Noisy Data for Fine-Grained Recognition
- Morphological Error Detection in 3D Segmentations
- MetaLabelNet: Learning to Generate Soft-Labels from Noisy-Labels
- Learning Deep Networks from Noisy Labels with Dropout Regularization
- How to Train Triplet Networks with 100K Identities?
- DC-Check: A Data-Centric AI checklist to guide the development of reliable machine learning systems
- Reinforcement Learning with Perturbed Rewards
- Deep Self-Learning From Noisy Labels
- Variational Rectification Inference for Learning with Noisy Labels
- Segmentation-Aware Image Denoising without Knowing True Segmentation
- Simple and Effective Regularization Methods for Training on Noisily Labeled Data with Generalization Guarantee
- ProSelfLC: Progressive Self Label Correction for Training Robust Deep Neural Networks
- Asymmetric Co-Teaching for Unsupervised Cross Domain Person Re-Identification
- Large Margin Deep Networks for Classification
- KNN-BERT: Fine-Tuning Pre-Trained Models with KNN Classifier
- Lung Cancer Screening Using Adaptive Memory-Augmented Recurrent Networks
- Breast Mass Detection with Faster R-CNN: On the Feasibility of Learning from Noisy Annotations
- RSSI-Based Hybrid Beamforming Design with Deep Learning
- Improving Medical Image Classification with Label Noise Using Dual-uncertainty Estimation
- Enhancing Robustness of On-line Learning Models on Highly Noisy Data
- A Semi-Supervised Two-Stage Approach to Learning from Noisy Labels
- Weak-shot Fine-grained Classification via Similarity Transfer
- In Defense of the Triplet Loss Again: Learning Robust Person Re-Identification with Fast Approximated Triplet Loss and Label Distillation
- Leveraging Multi-Source Weak Social Supervision for Early Detection of Fake News
- Learning with Biased Complementary Labels
- Identifying Mislabeled Images in Supervised Learning Utilizing Autoencoder
- Derivative Manipulation for General Example Weighting
- Supervised Contrastive Learning with Nearest Neighbor Search for Speech Emotion Recognition
- Robust Training with Ensemble Consensus
- Inferring Spatial Uncertainty in Object Detection
- SOSELETO: A Unified Approach to Transfer Learning and Training with Noisy Labels
- Task Agnostic Robust Learning on Corrupt Outputs by Correlation-Guided Mixture Density Networks
- Jo-SRC: A Contrastive Approach for Combating Noisy Labels
- Policy Learning Using Weak Supervision
- Learning from Indirect Observations
- Tag Prediction at Flickr: a View from the Darkroom
- "Is not the truth the truth?": Analyzing the Impact of User Validations for Bus In/Out Detection in Smartphone-based Surveys
- When Optimizing -divergence is Robust with Label Noise
- Improving Generalization by Controlling Label-Noise Information in Neural Network Weights
- Less is More: Learning Highlight Detection from Video Duration
- Robust and On-the-fly Dataset Denoising for Image Classification
- The Resistance to Label Noise in K-NN and DNN Depends on its Concentration
- Weakly Supervised Learning with Side Information for Noisy Labeled Images
- An Effective Label Noise Model for DNN Text Classification
- Attend in groups: a weakly-supervised deep learning framework for learning from web data
- Deep Learning from Small Amount of Medical Data with Noisy Labels: A Meta-Learning Approach
- Learning from Noisy Web Data with Category-level Supervision
- Deep Bilevel Learning
- Knodle: Modular Weakly Supervised Learning with PyTorch
- Learning from Web Data: the Benefit of Unsupervised Object Localization
- Learnability with Indirect Supervision Signals
- Contemplating Visual Emotions: Understanding and Overcoming Dataset Bias
- Learning with Out-of-Distribution Data for Audio Classification
- Robust Deep Graph Based Learning for Binary Classification
- Image recognition from raw labels collected without annotators
- Deep k-NN for Noisy Labels
- Joint Negative and Positive Learning for Noisy Labels
- Exponentiated Gradient Reweighting for Robust Training Under Label Noise and Beyond
- Co-matching: Combating Noisy Labels by Augmentation Anchoring
- KNN-enhanced Deep Learning Against Noisy Labels
- StackMix: A complementary Mix algorithm
- Learning from Multiple Annotators by Incorporating Instance Features
- Learning with Noisy Labels by Efficient Transition Matrix Estimation to Combat Label Miscorrection
- Noisy Annotation Refinement for Object Detection
- NLNL: Negative Learning for Noisy Labels
- Safeguarded Dynamic Label Regression for Generalized Noisy Supervision
- The Effects of Noisy Labels on Deep Convolutional Neural Networks for Music Tagging
- Video Representation Learning and Latent Concept Mining for Large-scale Multi-label Video Classification
- Avoidance of Manual Labeling in Robotic Autonomous Navigation Through Multi-Sensory Semi-Supervised Learning
- Combining Weakly and Webly Supervised Learning for Classifying Food Images
- Limited Gradient Descent: Learning With Noisy Labels
- Improving Medical Annotation Quality to Decrease Labeling Burden Using Stratified Noisy Cross-Validation
- Generative-Discriminative Complementary Learning
- Product Image Recognition with Guidance Learning and Noisy Supervision
- Adversarial-Based Knowledge Distillation for Multi-Model Ensemble and Noisy Data Refinement
- Detecting Fake News with Weak Social Supervision
- Negative-ResNet: Noisy Ambulatory Electrocardiogram Signal Classification Scheme
- From Recognition to Prediction: Analysis of Human Action and Trajectory Prediction in Video
- Facial Emotion Recognition with Noisy Multi-task Annotations
- Bootstrapping Deep Neural Networks from Approximate Image Processing Pipelines
- Matrix Smoothing: A Regularization for DNN with Transition Matrix under Noisy Labels
- RAD: On-line Anomaly Detection for Highly Unreliable Data
- Fine-grained Classification using Heterogeneous Web Data and Auxiliary Categories
- Deep Learning for Earth Image Segmentation based on Imperfect Polyline Labels with Annotation Errors
- A Survey on Deep Learning with Noisy Labels: How to train your model when you cannot trust on the annotations?
- Graph convolutional networks for learning with few clean and many noisy labels
- ARCA23K: An audio dataset for investigating open-set label noise
- Meta Label Correction for Noisy Label Learning
- Automatic Dataset Augmentation
- Training Classifiers that are Universally Robust to All Label Noise Levels
- Alternating Loss Correction for Preterm-Birth Prediction from EHR Data with Noisy Labels
- Robust Collaborative Learning with Noisy Labels
- UniMoCo: Unsupervised, Semi-Supervised and Full-Supervised Visual Representation Learning
- ADA: A Game-Theoretic Perspective on Data Augmentation for Object Detection
- Enhancing Segment-Based Speech Emotion Recognition by Deep Self-Learning
- An In-depth Analysis of Passage-Level Label Transfer for Contextual Document Ranking
- Self-Learning with Rectification Strategy for Human Parsing
- Towards Robust Pattern Recognition: A Review
- MetaXT: Meta Cross-Task Transfer between Disparate Label Spaces
- Towards Bottom-Up Analysis of Social Food
- Skeptical Deep Learning with Distribution Correction
- Probabilistic Decoupling of Labels in Classification
- Overcoming Conflicting Data when Updating a Neural Semantic Parser
- Training Convolutional Networks with Web Images
- Joint Text and Label Generation for Spoken Language Understanding
- A robust approach for deep neural networks in presence of label noise: relabelling and filtering instances during training
- Discovering Visual Concept Structure with Sparse and Incomplete Tags
- Analysis of classifiers robust to noisy labels
- Suppressing Mislabeled Data via Grouping and Self-Attention
- A Non-Intrusive Correction Algorithm for Classification Problems with Corrupted Data
- On tuning deep learning models: a data mining perspective
- Theoretical Guarantees of Deep Embedding Losses Under Label Noise
- Pseudo-positive regularization for deep person re-identification
- Task-Driven Data Verification via Gradient Descent
- Towards a New Understanding of the Training of Neural Networks with Mislabeled Training Data