mlsanityData to 7 Oct 2026

Robust Object Detection via Instance-Level Temporal Cycle Confusion

arXiv 2104.08381 ↗First posted 16 Apr 2021cs.CVcited by 24

Not on the map. Why?

Abstract (arXiv)

Building reliable object detectors that are robust to domain shifts, such as various changes in context, viewpoint, and object appearances, is critical for real-world applications. In this work, we study the effectiveness of auxiliary self-supervised tasks to improve the out-of-distribution generalization of object detectors. Inspired by the principle of maximum entropy, we introduce a novel self-supervised task, instance-level temporal cycle confusion (CycConf), which operates on the region features of the object detectors. For each object, the task is to find the most different object proposals in the adjacent frame in a video and then cycle back to itself for self-supervision. CycConf encourages the object detector to explore invariant structures across instances under various motions, which leads to improved model robustness in unseen domains at test time. We observe consistent out-of-domain performance improvements when training object detectors in tandem with self-supervised tasks on large-scale video datasets (BDD100K and Waymo open data). The joint training framework also establishes a new state-of-the-art on standard unsupervised domain adaptative detection benchmarks (Cityscapes, Foggy Cityscapes, and Sim10K). The code and models are available at https://github.com/xinw1012/cycle-confusion.

Key facts. Cited by 24 papers in the map (data through 7 Oct 2026). Builds on 39 of its 48 references that are in the map. First posted on arXiv 16 Apr 2021.

Builds on

39 of 48 references.

References · 39

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  3. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  4. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,617
  5. 2020End-to-End Object Detection with Transformerscited by 7,384
  6. 2016Feature Pyramid Networks for Object Detectioncited by 5,021
  7. 2016The Cityscapes Dataset for Semantic Urban Scene Understandingcited by 4,427
  8. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,857
  9. 2019Scalability in Perception for Autonomous Driving: Waymo Open Datasetcited by 2,436
  10. 2018Unsupervised Representation Learning by Predicting Image Rotationscited by 1,586
  11. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,430
  12. 2016Context Encoders: Feature Learning by Inpaintingcited by 1,424
27 more
  1. 2016Unsupervised Learning of Visual Representations by Solving Jigsaw Puzzlescited by 1,379
  2. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,366
  3. 2018BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learningcited by 1,282
  4. 2015Unsupervised Visual Representation Learning by Context Predictioncited by 1,062
  5. 2019AugMix: A Simple Data Processing Method to Improve Robustness and Uncertaintycited by 897
  6. 2019Using Self-Supervised Learning Can Improve Model Robustness and Uncertaintycited by 511
  7. 2017Time-Contrastive Networks: Self-Supervised Learning from Videocited by 453
  8. 2018Domain Adaptive Faster R-CNN for Object Detection in the Wildcited by 442
  9. 2017Semantic Foggy Scene Understanding with Synthetic Datacited by 401
  10. 2019Using Pre-Training Can Improve Model Robustness and Uncertaintycited by 373
  11. 2018Strong-Weak Distribution Alignment for Adaptive Object Detectioncited by 269
  12. 2019Learning Correspondence from the Cycle-Consistency of Timecited by 269
  13. 2016Split-Brain Autoencoders: Unsupervised Learning by Cross-Channel Predictioncited by 217
  14. 2018Do CIFAR-10 Classifiers Generalize to CIFAR-10?cited by 197
  15. 2016Driving in the Matrix: Can Virtual Worlds Replace Human-Generated Annotations for Real World Tasks?cited by 194
  16. 2021Maximum Entropy RL (Provably) Solves Some Robust RL Problemscited by 169
  17. 2019Temporal Cycle-Consistency Learningcited by 151
  18. 2019Exploring Object Relation in Mean Teacher for Cross-Domain Detectioncited by 136
  19. 2020Exploring Categorical Regularization for Domain Adaptive Object Detectioncited by 130
  20. 2019Diversify and Match: A Domain Adaptive Representation Learning Paradigm for Object Detectioncited by 122
  21. 2019Unsupervised Domain Adaptation through Self-Supervisioncited by 116
  22. 2019Multi-adversarial Faster-RCNN for Unrestricted Object Detectioncited by 111
  23. 2020Cross-domain Detection via Graph-induced Prototype Alignmentcited by 109
  24. 2020Every Pixel Matters: Center-aware Feature Alignment for Domain Adaptive Object Detectorcited by 106
  25. 2016Ambient Sound Provides Supervision for Visual Learningcited by 77
  26. 2020Adapting Object Detectors with Conditional Domain Normalizationcited by 55
  27. 2020Collaborative Training between Region Proposal Localization and Classification for Domain Adaptive Object Detectioncited by 44

Built on by

Cited by year

Citing papers per year: 2021: 1; 2022: 6; 2023: 11; 2024: 2; 2025: 4.

24 papers.

Sort:

  1. 2022VOS: Learning What You Don't Know by Virtual Outlier Synthesiscited by 300
  2. 2021Unsupervised Domain Adaptation of Object Detectors: A Surveycited by 93
  3. 2022Unknown-Aware Object Detection: Learning What You Don't Know from Videos in the Wildcited by 70
  4. 2022PASTA: Proportional Amplitude Spectrum Training Augmentation for Syn-to-Real Domain Generalizationcited by 38
  5. 2023Object-Aware Domain Generalization for Object Detectioncited by 21
  6. 2023Towards Building Self-Aware Object Detectors via Reliable Uncertainty Quantification and Calibrationcited by 19
  7. 2023Normalizing Flow based Feature Synthesis for Outlier-Aware Object Detectioncited by 16
  8. 2024PhysAug: A Physical-guided and Frequency-based Data Augmentation for Single-Domain Generalized Object Detectioncited by 11
  9. 2024Can OOD Object Detectors Learn from Foundation Models?cited by 10
  10. 2023Bridging the Sim2Real gap with CARE: Supervised Detection Adaptation with Conditional Alignment and Reweightingcited by 9
  11. 2022Normalization Perturbation: A Simple Domain Generalization Method for Real-World Domain Shiftscited by 8
  12. 2023NormAUG: Normalization-guided Augmentation for Domain Generalizationcited by 8
12 more
  1. 2022Composite Learning for Robust and Effective Dense Predictionscited by 4
  2. 2022Visual Attention Emerges from Recurrent Sparse Reconstructioncited by 3
  3. 2025DG-DETR: Toward Domain Generalized Detection Transformercited by 3
  4. 2023Mind the Backbone: Minimizing Backbone Distortion for Robust Object Detectioncited by 2
  5. 2023DoubleAUG: Single-domain Generalized Object Detector in Urban via Color Perturbation and Dual-style Memorycited by 2
  6. 2023Drawing the Same Bounding Box Twice? Coping Noisy Annotations in Object Detection with Repeated Labelscited by 1
  7. 2025Exploring Syn-to-Real Domain Adaptation for Military Target Detectioncited by 1
  8. 2023Patch-aware Batch Normalization for Improving Cross-domain Robustnesscited by 0
  9. 2023Seeking Flat Minima with Mean Teacher on Semi- and Weakly-Supervised Domain Generalization for Object Detectioncited by 0
  10. 2023Proposal-Level Unsupervised Domain Adaptation for Open World Unbiased Detectorcited by 0
  11. 2025Boosting Single-domain Generalized Object Detection via Vision-Language Knowledge Interactioncited by 0
  12. 2025De-Simplifying Pseudo Labels to Enhancing Domain Adaptive Object Detectioncited by 0