Nuisance-Label Supervision: Robustness Improvement by Free Labels
Not on the map. Why?
Abstract (arXiv)
In this paper, we present a Nuisance-label Supervision (NLS) module, which can make models more robust to nuisance factor variations. Nuisance factors are those irrelevant to a task, and an ideal model should be invariant to them. For example, an activity recognition model should perform consistently regardless of the change of clothes and background. But our experiments show existing models are far from this capability. So we explicitly supervise a model with nuisance labels to make extracted features less dependent on nuisance factors. Although the values of nuisance factors are rarely annotated, we demonstrate that besides existing annotations, nuisance labels can be acquired freely from data augmentation and synthetic data. Experiments show consistent improvement in robustness towards image corruption and appearance change in action recognition.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 33 of its 41 references that are in the map. First posted on arXiv 14 Oct 2021.
Builds on
33 of 41 references.
References · 33
- 2014Microsoft COCO: Common Objects in Contextcited by 17,608
- 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
- 2017Quo Vadis, Action Recognition? A New Model and the Kinetics Datasetcited by 3,270
- 2015Learning Deep Features for Discriminative Localizationcited by 2,436
- 2014Unsupervised Domain Adaptation by Backpropagationcited by 1,917
- 2017Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real Worldcited by 1,498
- 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,432
- 2016Temporal Segment Networks: Towards Good Practices for Deep Action Recognitioncited by 902
- 2019AugMix: A Simple Data Processing Method to Improve Robustness and Uncertaintycited by 901
- 2016Playing for Data: Ground Truth from Computer Gamescited by 818
- 2016NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysiscited by 686
- 2018Taskonomy: Disentangling Task Transfer Learningcited by 634
21 more
- 2014Deeply-Supervised Netscited by 367
- 2017Learning from Synthetic Humanscited by 359
- 2018Learning Not to Learn: Training Deep Neural Networks with Biased Datacited by 255
- 2019MNIST-C: A Robustness Benchmark for Computer Visioncited by 138
- 2019Why Can't I Dance in the Mall? Learning to Mitigate Scene Bias in Action Recognitioncited by 117
- 2018Turning a Blind Eye: Explicit Removal of Biases and Variation from Deep Neural Network Embeddingscited by 114
- 20182D/3D Pose Estimation and Action Recognition using Multitask Deep Learningcited by 103
- 2020Don't Judge an Object by Its Context: Learning to Overcome Contextual Biascited by 91
- 2017Cross-Domain Self-supervised Multi-task Feature Learning using Synthetic Imagerycited by 56
- 2016Synthesizing Training Images for Boosting Human 3D Pose Estimationcited by 52
- 2019Synthetic Humans for Action Recognition from Unseen Viewpointscited by 46
- 2016MoCap-guided Data Augmentation for 3D Pose Estimation in the Wildcited by 37
- 2018Glimpse Clouds: Human Activity Recognition from Unstructured Feature Pointscited by 33
- 2016Procedural Generation of Videos to Train Deep Action Recognition Networkscited by 31
- 2020A simple way to make neural networks robust against diverse image corruptionscited by 30
- 2016Deep Supervision with Shape Concepts for Occlusion-Aware 3D Object Parsingcited by 19
- 2014R-CNNs for Pose Estimation and Action Detectioncited by 12
- 2017Pose-conditioned Spatio-Temporal Attention for Human Action Recognitioncited by 4
- 2018Image-based Synthesis for Deep 3D Human Pose Estimationcited by 3
- 2019RSA: Randomized Simulation as Augmentation for Robust Human Action Recognitioncited by 2
- 2019Identity Preserve Transform: Understand What Activity Classification Models Have Learntcited by 2
Built on by
Citing papers per year: 2025: 1.
1 paper.