mlsanityData to 14 Oct 2026

MUM : Mix Image Tiles and UnMix Feature Tiles for Semi-Supervised Object Detection

arXiv 2111.10958 ↗First posted 22 Nov 2021cs.CVcited by 21

On the map: Knowledge distillation → Detection and dense prediction distillation

Abstract (arXiv)

Many recent semi-supervised learning (SSL) studies build teacher-student architecture and train the student network by the generated supervisory signal from the teacher. Data augmentation strategy plays a significant role in the SSL framework since it is hard to create a weak-strong augmented input pair without losing label information. Especially when extending SSL to semi-supervised object detection (SSOD), many strong augmentation methodologies related to image geometry and interpolation-regularization are hard to utilize since they possibly hurt the location information of the bounding box in the object detection task. To address this, we introduce a simple yet effective data augmentation method, Mix/UnMix (MUM), which unmixes feature tiles for the mixed image tiles for the SSOD framework. Our proposed method makes mixed input image tiles and reconstructs them in the feature space. Thus, MUM can enjoy the interpolation-regularization effect from non-interpolated pseudo-labels and successfully generate a meaningful weak-strong pair. Furthermore, MUM can be easily equipped on top of various SSOD methods. Extensive experiments on MS-COCO and PASCAL VOC datasets demonstrate the superiority of MUM by consistently improving the mAP performance over the baseline in all the tested SSOD benchmark protocols.

Key facts. Cited by 21 papers in the map (data through 14 Oct 2026). Placed in Detection and dense prediction distillation, one of 12 directions in Knowledge distillation. Builds on 34 of its 46 references that are in the map. First posted on arXiv 22 Nov 2021.

Builds on

34 of 46 references.

References · 34

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  3. 2021Swin Transformer: Hierarchical Vision Transformer using Shifted Windowscited by 12,139
  4. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,639
  5. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  6. 2017mixup: Beyond Empirical Risk Minimizationcited by 5,183
  7. 2019CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Featurescited by 2,884
  8. 2019RandAugment: Practical automated data augmentation with a reduced search spacecited by 2,269
  9. 2020FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidencecited by 2,085
  10. 2020YOLOv4: Optimal Speed and Accuracy of Object Detectioncited by 1,630
  11. 2017Improved Regularization of Convolutional Neural Networks with Cutoutcited by 1,628
  12. 2019MixMatch: A Holistic Approach to Semi-Supervised Learningcited by 1,391
22 more
  1. 2019Self-training with Noisy Student improves ImageNet classificationcited by 1,311
  2. 2021YOLOX: Exceeding YOLO Series in 2021cited by 1,071
  3. 2019Unsupervised Data Augmentation for Consistency Trainingcited by 1,060
  4. 2017Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learningcited by 920
  5. 2017Focal Loss for Dense Object Detectioncited by 752
  6. 2019Pseudo-Labeling and Confirmation Bias in Deep Semi-Supervised Learningcited by 510
  7. 2017Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning resultscited by 472
  8. 2019ReMixMatch: Semi-Supervised Learning with Distribution Alignment and Augmentation Anchoringcited by 349
  9. 2021Unbiased Teacher for Semi-Supervised Object Detectioncited by 306
  10. 2015What's the Point: Semantic Segmentation with Point Supervisioncited by 298
  11. 2019Label Propagation for Deep Semi-supervised Learningcited by 289
  12. 2020A Simple Semi-Supervised Learning Framework for Object Detectioncited by 282
  13. 2020Puzzle Mix: Exploiting Saliency and Local Statistics for Optimal Mixupcited by 261
  14. 2019Interpolation Consistency Training for Semi-Supervised Learningcited by 251
  15. 2014Learning with Pseudo-Ensemblescited by 241
  16. 2021Co-Mixup: Saliency Guided Joint Mixup with Supermodular Diversitycited by 128
  17. 2021Instant-Teaching: An End-to-End Semi-Supervised Object Detection Frameworkcited by 113
  18. 2018Understanding and Improving Interpolation in Autoencoders via an Adversarial Regularizercited by 88
  19. 2017Between-class Learning for Image Classificationcited by 50
  20. 2021Data-Uncertainty Guided Multi-Phase Learning for Semi-Supervised Object Detectioncited by 49
  21. 2020Interpolation-based semi-supervised learning for object detectioncited by 41
  22. 2020Structured Consistency Loss for semi-supervised semantic segmentationcited by 40

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2022: 7; 2023: 7; 2024: 5; 2025: 1; 2026: 1.

21 papers.

Sort:

By area:

  1. 2023Bidirectional Copy-Paste for Semi-Supervised Medical Image Segmentationcited by 116
  2. 2024A Survey on Mixup Augmentations and Beyondcited by 20
  3. 2022DTG-SSOD: Dense Teacher Guidance for Semi-Supervised Object Detectioncited by 12
  4. 2022Dynamic Curriculum Learning for Great Ape Detection in the Wildcited by 10
  5. 2023Virtual Category Learning: A Semi-Supervised Learning Method for Dense Prediction with Extremely Limited Labelscited by 10
  6. 2022Cycle Self-Training for Semi-Supervised Object Detection with Distribution Consistency Reweightingcited by 5
  7. 2022Semi-Supervised Object Detection with Object-wise Contrastive Learning and Regression Uncertaintycited by 5
  8. 2023Semi-supervised Object Detection: A Survey on Recent Research and Progresscited by 5
  9. 2022Adaptive Self-Training for Object Detectioncited by 4
  10. 2024Semi-Supervised Object Detection: A Survey on Progress from CNN to Transformercited by 4
  11. 2022Pose-MUM : Reinforcing Key Points Relationship for Semi-Supervised Human Pose Estimationcited by 3
  12. 2022Dual-Curriculum Teacher for Domain-Inconsistent Object Detection in Autonomous Drivingcited by 2
9 more
  1. 2023Infinite Class Mixupcited by 2
  2. 2024Semi-Supervised Unconstrained Head Pose Estimation in the Wildcited by 2
  3. 2023Training-based Model Refinement and Representation Disagreement for Semi-Supervised Object Detectioncited by 1
  4. 2024IMWA: Iterative Model Weight Averaging Benefits Class-Imbalanced Learning Taskscited by 1
  5. 2023Low-Confidence Samples Mining for Semi-supervised Object Detectioncited by 0
  6. 2023Rethinking Scale Imbalance in Semi-supervised Object Detection for Aerial Imagescited by 0
  7. 2024Collaboration of Teachers for Semi-supervised Object Detectioncited by 0
  8. 2025P3Net: Progressive and Periodic Perturbation for Semi-Supervised Medical Image Segmentationcited by 0
  9. 2026DDS-UDA: Dual-Domain Synergy for Unsupervised Domain Adaptation in Joint Segmentation of Optic Disc and Optic Cupcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2023Hierarchical Supervision and Shuffle Data Augmentation for 3D Semi-Supervised Object Detection
  2. 2021Humble Teachers Teach Better Students for Semi-Supervised Object Detection
  3. 2024Co-Learning: Towards Semi-Supervised Object Detection with Road-side Cameras
  4. 2023Towards End-to-end Semi-supervised Learning for One-stage Object Detection
  5. 2022Boosting Semi-Supervised 3D Object Detection with Semi-Sampling
  6. 2023Active Teacher for Semi-Supervised Object Detection
  7. 2024Semi-supervised 3D Object Detection with PatchTeacher and PillarMix
  8. 2024Semi-Supervised 3D Object Detection with Channel Augmentation using Transformation Equivariance