目标检测中的不平衡问题及解决方案论文汇总

以你之姓@ 2023-07-02 09:27 14阅读 0赞

转载自[https://github.com/kemaloksuz/ObjectDetectionImbalance][https_github.com_kemaloksuz_ObjectDetectionImbalance]

# A Repository of the Papers Addressing Imbalance Problems in Object Detection #

This repository provides an up-to-date the list of studies addressing imbalance problems in object detection. It follows the taxonomy provided in the following paper(please cite the paper if you benefit from this repository):

K. Oksuz, B. C. Cam, S. Kalkan, E. Akbas, “Imbalance Problems in Object Detection: A Review”, (under review), 2019.[\[preprint\]][preprint]

# Table of Contents (Follows the taxonomy in the paper) #

1.  Class Imbalance  
    1.1 Foreground-Backgorund Class Imbalance  
    1.2 Foreground-Foreground Class Imbalance
2.  Scale Imbalance  
    2.1 Object/box-level Scale Imbalance  
    2.2 Feature-level Imbalance
3.  Spatial Imbalance  
    3.1 Imbalance in Regression Loss  
    3.2 IoU Distribution Imbalance  
    3.3 Object Location Imbalance
4.  Objective Imbalance

# 1. Class Imbalance #

## 1.1. Foreground-Backgorund Class Imbalance ##

*  Hard Sampling Methods
    
     *  Random Sampling
     *  Hard Example Mining
        
         *  Bootstrapping, NeurIPS 1996, [\[paper\]][paper]
         *  SSD, ECCV 2016, [\[paper\]][paper 1]
         *  Online Hard Example Mining, CVPR 2016, [\[paper\]][paper 2]
         *  IoU-based Sampling, CVPR 2019, [\[paper\]][paper 3]
     *  Limit Search Space
        
         *  Two-stage Object Detectors
         *  IoU-lower Bound, ICCV 2015, [\[paper\]][paper 4]
         *  Objectness Prior, CVPR 2017, [\[paper\]][paper 5]
         *  Negative Anchor Filtering, CVPR 2018, [\[paper\]][paper 6]
         *  Objectness Module, ICCV 2019, [\[paper\]][paper 7]
 *  Soft Sampling Methods
    
     *  Focal Loss, ICCV 2017, [\[paper\]][paper 8]
     *  Gradient Harmonizing Mechanism, AAAI 2019, [\[paper\]][paper 9]
     *  Prime Sample Attention, arXiv 2019, [\[paper\]][paper 10]
 *  Sampling-Free Methods
    
     *  Is Sampling Heuristics Necessary in Training Deep Object Detectors?, arXiv 2019, [\[paper\]][paper 11]
     *  Residual Objectness for Imbalance Reduction, arXiv 2019, [\[paper\]][paper 12]
     *  AP Loss, CVPR 2019, [\[paper\]][paper 13]
     *  DR Loss, arXiv 2019, [\[paper\]][paper 14]
 *  Generative Methods
    
     *  Adversarial Faster-RCNN, CVPR 2017, [\[paper\]][paper 15]
     *  Task Aware Data Synthesis, CVPR 2019, [\[paper\]][paper 16]
     *  PSIS, arXiv 2019, [\[paper\]][paper 17]
     *  Bounding Box Generator, WACV 2020, [\[paper\]][paper 18]

## 1.2. Foreground-Foreground Class Imbalance ##

*  Fine-tuning Long Tail Distribution for Obj.Det., CVPR 2016, [\[paper\]][paper 19]
 *  PSIS, arXiv 2019, [\[paper\]][paper 17]
 *  OFB Sampling, WACV 2020, [\[paper\]][paper 18]

# 2. Scale Imbalance #

## 2.1. Object/box-level Scale Imbalance ##

*  Methods Predicting from the Feature Hierarchy of Backbone Features
    
     *  Scale-dependent Pooling, CVPR 2016, [\[paper\]][paper 20]
     *  SSD, ECCV 2016, [\[paper\]][paper 1]
     *  Multi Scale CNN, ECCV 2016, [\[paper\]][paper 21]
     *  Scale Aware Fast R-CNN, IEEE Transactions on Multimedia, 2018 [\[paper\]][paper 22]
 *  Methods Based on Feature Pyramids
    
     *  FPN, CVPR 2017, [\[paper\]][paper 23]
     *  See feature-level imbalance methods
 *  Methods Based on Image Pyramids
    
     *  SNIP, CVPR 2018, [\[paper\]][paper 24]
     *  SNIPER, NeurIPS 2018, [\[paper\]][paper 25]
 *  Methods Combining Image and Feature Pyramids
    
     *  Efficient Featurized Image Pyramids, CVPR 2019, [\[paper\]][paper 26]
     *  Enriched Feature Guided Refinement Network, ICCV 2019, [\[paper\]][paper 7]
     *  Super-Resolution for Small Objects, ICCV 2019, [\[paper\]][paper 27]
     *  Scale Aware Trident Network, ICCV 2019, [\[paper\]][paper 28]

## 2.2. Feature-level Imbalance ##

*  Methods Using Pyramidal Features as a Basis
    
     *  PANet, CVPR 2018, [\[paper\]][paper 29]
     *  Libra FPN, CVPR 2019, [\[paper\]][paper 3]
 *  Methods Using Backbone Features as a Basis
    
     *  STDN, CVPR 2018, [\[paper\]][paper 30]
     *  Parallel-FPN, ECCV 2018, [\[paper\]][paper 31]
     *  Deep Feature Pyramid Reconfiguration, ECCV 2018, [\[paper\]][paper 32]
     *  Zoom Out-and-In, IJCV 2019, [\[paper\]][paper 33]
     *  Multi-level FPN, AAAI 2019, [\[paper\]][paper 34]
     *  NAS-FPN, CVPR 2019, [\[paper\]][paper 35]
     *  Auto-FPN, ICCV 2019, [\[paper\]][paper 36]

# 3. Spatial Imbalance #

## 3.1. Imbalance in Regression Loss ##

*  Lp norm based
    
     *  Smooth L1, ICCV 2015, [\[paper\]][paper 4]
     *  Balanced L1, CVPR 2019, [\[paper\]][paper 37]
     *  KL Loss, CVPR 2019, [\[paper\]][paper 38]
     *  Gradient Harmonizing Mechanism, AAAI 2019, [\[paper\]][paper 39]
 *  IoU based
    
     *  IoU Loss, ACM IMM 2016, [\[paper\]][paper 40]
     *  Bounded IoU Loss, CVPR 2018, [\[paper\]][paper 41]
     *  Generalized IoU Loss, CVPR 2019, [\[paper\]][paper 42]
     *  Distance IoU Loss, AAAI 2020, [\[paper\]][paper 43]
     *  Complete IoU Loss, AAAI 2020, [\[paper\]][paper 43]

## 3.2. IoU Distribution Imbalance ##

*  Cascade R-CNN, CVPR 2018, [\[paper\]][paper 44]
 *  Hierarchical Shot Detector, ICCV 2019, [\[paper\]][paper 45]

## 3.3. Object Location Imbalance ##

*  Guided Anchoring, CVPR 2019, [\[paper\]][paper 46]
 *  FreeAnchor, NeurIPS 2019, [\[paper\]][paper 47]

# 4. Objective Imbalance #

*  Task Weighting
 *  Classification Aware Regression Loss, arXiv 2019, [\[paper\]][paper 10]
 *  Guided Loss, arXiv 2019, [\[paper\]][paper 11]

[https_github.com_kemaloksuz_ObjectDetectionImbalance]: https://github.com/kemaloksuz/ObjectDetectionImbalance
[preprint]: https://arxiv.org/abs/1909.00169
[paper]: https://papers.nips.cc/paper/1168-human-face-detection-in-visual-scenes.pdf
[paper 1]: http://www.cs.unc.edu/~wliu/papers/ssd.pdf
[paper 2]: https://zpascal.net/cvpr2016/Shrivastava_Training_Region-Based_Object_CVPR_2016_paper.pdf
[paper 3]: http://openaccess.thecvf.com/content_CVPR_2019/papers/Pang_Libra_R-CNN_Towards_Balanced_Learning_for_Object_Detection_CVPR_2019_paper.pdf
[paper 4]: https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Girshick_Fast_R-CNN_ICCV_2015_paper.pdf
[paper 5]: http://zpascal.net/cvpr2017/Kong_RON_Reverse_Connection_CVPR_2017_paper.pdf
[paper 6]: http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhang_Single-Shot_Refinement_Neural_CVPR_2018_paper.pdf
[paper 7]: http://openaccess.thecvf.com/content_ICCV_2019/papers/Nie_Enriched_Feature_Guided_Refinement_Network_for_Object_Detection_ICCV_2019_paper.pdf
[paper 8]: http://openaccess.thecvf.com/content_ICCV_2017/papers/Lin_Focal_Loss_for_ICCV_2017_paper.pdf
[paper 9]: https://aaai.org/ojs/index.php/AAAI/article/view/4877
[paper 10]: https://arxiv.org/pdf/1904.04821.pdf
[paper 11]: https://arxiv.org/pdf/1909.04868.pdf
[paper 12]: https://arxiv.org/pdf/1908.09075.pdf
[paper 13]: http://openaccess.thecvf.com/content_CVPR_2019/papers/Chen_Towards_Accurate_One-Stage_Object_Detection_With_AP-Loss_CVPR_2019_paper.pdf
[paper 14]: https://arxiv.org/pdf/1907.10156.pdf
[paper 15]: http://zpascal.net/cvpr2017/Wang_A-Fast-RCNN_Hard_Positive_CVPR_2017_paper.pdf
[paper 16]: http://openaccess.thecvf.com/content_CVPR_2019/papers/Tripathi_Learning_to_Generate_Synthetic_Data_via_Compositing_CVPR_2019_paper.pdf
[paper 17]: https://arxiv.org/pdf/1906.00358.pdf
[paper 18]: https://arxiv.org/abs/1909.09777
[paper 19]: http://openaccess.thecvf.com/content_cvpr_2016/papers/Ouyang_Factors_in_Finetuning_CVPR_2016_paper.pdf
[paper 20]: http://openaccess.thecvf.com/content_cvpr_2016/papers/Yang_Exploit_All_the_CVPR_2016_paper.pdf
[paper 21]: https://arxiv.org/pdf/1607.07155.pdf
[paper 22]: https://ieeexplore.ieee.org/document/8060595
[paper 23]: https://zpascal.net/cvpr2017/Lin_Feature_Pyramid_Networks_CVPR_2017_paper.pdf
[paper 24]: http://openaccess.thecvf.com/content_cvpr_2018/papers/Singh_An_Analysis_of_CVPR_2018_paper.pdf
[paper 25]: https://papers.nips.cc/paper/8143-sniper-efficient-multi-scale-training.pdf
[paper 26]: http://openaccess.thecvf.com/content_CVPR_2019/papers/Pang_Efficient_Featurized_Image_Pyramid_Network_for_Single_Shot_Detector_CVPR_2019_paper.pdf
[paper 27]: http://openaccess.thecvf.com/content_ICCV_2019/papers/Noh_Better_to_Follow_Follow_to_Be_Better_Towards_Precise_Supervision_ICCV_2019_paper.pdf
[paper 28]: http://openaccess.thecvf.com/content_ICCV_2019/papers/Li_Scale-Aware_Trident_Networks_for_Object_Detection_ICCV_2019_paper.pdf
[paper 29]: http://openaccess.thecvf.com/content_cvpr_2018/papers/Liu_Path_Aggregation_Network_CVPR_2018_paper.pdf
[paper 30]: http://openaccess.thecvf.com/content_cvpr_2018/CameraReady/1376.pdf
[paper 31]: http://openaccess.thecvf.com/content_ECCV_2018/papers/Seung-Wook_Kim_Parallel_Feature_Pyramid_ECCV_2018_paper.pdf
[paper 32]: https://eccv2018.org/openaccess/content_ECCV_2018/papers/Tao_Kong_Deep_Feature_Pyramid_ECCV_2018_paper.pdf
[paper 33]: https://arxiv.org/pdf/1709.04347.pdf
[paper 34]: https://arxiv.org/pdf/1811.04533.pdf
[paper 35]: http://openaccess.thecvf.com/content_CVPR_2019/papers/Ghiasi_NAS-FPN_Learning_Scalable_Feature_Pyramid_Architecture_for_Object_Detection_CVPR_2019_paper.pdf
[paper 36]: http://openaccess.thecvf.com/content_ICCV_2019/papers/Xu_Auto-FPN_Automatic_Network_Architecture_Adaptation_for_Object_Detection_Beyond_Classification_ICCV_2019_paper.pdf
[paper 37]: http://openaccess.thecvf.com/content_CVPR_2019/html/Pang_Libra_R-CNN_Towards_Balanced_Learning_for_Object_Detection_CVPR_2019_paper.html
[paper 38]: http://openaccess.thecvf.com/content_CVPR_2019/papers/He_Bounding_Box_Regression_With_Uncertainty_for_Accurate_Object_Detection_CVPR_2019_paper.pdf
[paper 39]: https://aaai.org/ojs/index.php/AAAI/article/view/4877/4750
[paper 40]: https://arxiv.org/pdf/1608.01471.pdf
[paper 41]: http://openaccess.thecvf.com/content_cvpr_2018/CameraReady/0794.pdf
[paper 42]: http://openaccess.thecvf.com/content_CVPR_2019/papers/Rezatofighi_Generalized_Intersection_Over_Union_A_Metric_and_a_Loss_for_CVPR_2019_paper.pdf
[paper 43]: https://arxiv.org/pdf/1911.08287.pdf
[paper 44]: http://openaccess.thecvf.com/content_cvpr_2018/papers/Cai_Cascade_R-CNN_Delving_CVPR_2018_paper.pdf
[paper 45]: http://openaccess.thecvf.com/content_ICCV_2019/papers/Cao_Hierarchical_Shot_Detector_ICCV_2019_paper.pdf
[paper 46]: http://openaccess.thecvf.com/content_CVPR_2019/html/Wang_Region_Proposal_by_Guided_Anchoring_CVPR_2019_paper.html
[paper 47]: https://papers.nips.cc/paper/8309-freeanchor-learning-to-match-anchors-for-visual-object-detection.pdf