Prime Sample Attention in Object Detection

@article{Cao2020PrimeSA,
  title={Prime Sample Attention in Object Detection},
  author={Yuhang Cao and Kai Chen and Chen Change Loy and Dahua Lin},
  journal={2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2020},
  pages={11580-11588}
}
  • Yuhang Cao, Kai Chen, +1 author Dahua Lin
  • Published 2020
  • Computer Science
  • 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
It is a common paradigm in object detection frameworks to treat all samples equally and target at maximizing the performance on average. In this work, we revisit this paradigm through a careful study on how different samples contribute to the overall performance measured in terms of mAP. Our study suggests that the samples in each mini-batch are neither independent nor equally important, and therefore a better classifier on average does not necessarily result in higher mAP. Motivated by this… Expand
Learning a Unified Sample Weighting Network for Object Detection
TLDR
A general loss function is devised to cover most region-based object detectors with various sampling strategies, and then a unified sample weighting network is proposed to predict a sample's task weights to consistently boost the detection accuracy. Expand
Are Sampling Heuristics Necessary in Object Detectors?
TLDR
This paper's investigation reveals that, with careful training and inference schemes, the well-known RetinaNet could still achieve similar accuracy even without Focal Loss, and proposes Sampling-Free mechanism as an alternative to sampling heuristics, which addresses the imbalance from aspects of initialization, loss and inference, thus avoiding laborious hyper-parameters tuning. Expand
IoU Regression with H+L-Sampling for Accurate Detection Confidence
  • Dong Wang, Huaming Wu
  • Computer Science, Medicine
  • Sensors
  • 2021
TLDR
An H+L-Sampling strategy, selecting the high and low IoU samples simultaneously, to effectively and simply train the branch of quality estimation, which inherits the effectiveness of consistent sampling and reduces the training difficulty of uniform sampling. Expand
Is Sampling Heuristics Necessary in Training Deep Object Detectors
TLDR
This paper proposes a guided loss scaling technique to control the classification loss during training, without using any hyper-parameter, and proposes an adaptive thresholding technique to refine predictions during inference, which constitute the Sampling-Free mechanism. Expand
Overlap Sampler for Region-Based Object Detection
TLDR
This work proposes Overlap Sampler that selects examples according to the overlaps among examples, which enables the training to focus on the important examples and yields considerable improvements for the instance segmentation task. Expand
Mutual Supervision for Dense Object Detection
  • Ziteng Gao, Limin Wang, Gangshan Wu
  • Computer Science
  • ArXiv
  • 2021
TLDR
A novel supervisory paradigm, termed as Mutual Supervision (MuSu), is explored to respectively and mutually assign training samples for the classification and regression head to ensure consistency in dense detectors. Expand
Resisting Crowd Occlusion and Hard Negatives for Pedestrian Detection in the Wild.
TLDR
A novel coulomb loss as a regulator on bounding box regression is designed to address the crowd occlusion and hard negatives, and an efficient semantic-driven strategy for selecting anchor locations is proposed which can sample informative negative examples at training phase for classification refinement. Expand
Dynamic Multi-Scale Loss Optimization for Object Detection
  • Yihao Luo, Xiang Cao, Juntao Zhang, Peng Cheng, Tianjiang Wang, Qi Feng
  • Computer Science
  • ArXiv
  • 2021
TLDR
This paper proposes an Adaptive Variance Weighting (AVW) to balance multi-scale loss according to the statistical variance, and develops a novel Reinforcement Learning Optimization (RLO) to decide the weighting scheme probabilistically during training. Expand
TOOD: Task-aligned One-stage Object Detection
  • Chengjian Feng, Yujie Zhong, Yu Gao, Matthew R. Scott, Weilin Huang
  • Computer Science
  • ArXiv
  • 2021
TLDR
This work proposes a Task-aligned One-stage Object Detection (TOOD) that explicitly aligns the two tasks in a learning-based manner, and designs a novel Task- aligned Head (T-Head) which offers a better balance between learning task-interactive and task-specific features, as well as a greater flexibility to learn the alignment via a task-aligned predictor. Expand
RankDetNet: Delving Into Ranking Constraints for Object Detection
TLDR
This work proposes a ranking-based optimization algorithm for harmoniously learning to rank and localize proposals in lieu of the classification task, and comprehensively investigates three types of ranking constraints, i.e., global ranking, class-specific ranking and IoU-guided ranking losses. Expand
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 37 REFERENCES
Training Region-Based Object Detectors with Online Hard Example Mining
TLDR
OHEM is a simple and intuitive algorithm that eliminates several heuristics and hyperparameters in common use that leads to consistent and significant boosts in detection performance on benchmarks like PASCAL VOC 2007 and 2012. Expand
Focal Loss for Dense Object Detection
TLDR
This paper proposes to address the extreme foreground-background class imbalance encountered during training of dense detectors by reshaping the standard cross entropy loss such that it down-weights the loss assigned to well-classified examples, and develops a novel Focal Loss, which focuses training on a sparse set of hard examples and prevents the vast number of easy negatives from overwhelming the detector during training. Expand
Focal Loss for Dense Object Detection
TLDR
This paper proposes to address the extreme foreground-background class imbalance encountered during training of dense detectors by reshaping the standard cross entropy loss such that it down-weights the loss assigned to well-classified examples, and develops a novel Focal Loss, which focuses training on a sparse set of hard examples and prevents the vast number of easy negatives from overwhelming the detector during training. Expand
Libra R-CNN: Towards Balanced Learning for Object Detection
TLDR
Libra R-CNN is proposed, a simple but effective framework towards balanced learning for object detection that integrates three novel components: IoU-balanced sampling, balanced feature pyramid, and balanced L1 loss, respectively for reducing the imbalance at sample, feature, and objective level. Expand
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation
TLDR
This paper proposes a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 30% relative to the previous best result on VOC 2012 -- achieving a mAP of 53.3%. Expand
Cascade R-CNN: Delving Into High Quality Object Detection
TLDR
A simple implementation of the Cascade R-CNN is shown to surpass all single-model object detectors on the challenging COCO dataset, and experiments show that it is widely applicable across detector architectures, achieving consistent gains independently of the baseline detector strength. Expand
DR Loss: Improving Object Detection by Distributional Ranking
TLDR
This work proposes a novel distributional ranking (DR) loss, which mitigates the intra-class imbalance issue in background candidates but also improves the efficiency for the ranking algorithm. Expand
ScratchDet: Exploring to Train Single-Shot Object Detectors from Scratch
TLDR
The ScratchDet achieves the state-of-the-art accuracy on PASCAL VOC 2007, 2012 and MS COCO among all the train-from-scratch detectors and even performs better than several one-stage pretrained methods. Expand
Deep Regionlets for Object Detection
TLDR
This paper proposes a "region selection network" and a "gating network" that serves as a guidance on where to select regions to learn the features from in Regionlet and achieves comparable state-of-the-art results. Expand
Feature Pyramid Networks for Object Detection
TLDR
This paper exploits the inherent multi-scale, pyramidal hierarchy of deep convolutional networks to construct feature pyramids with marginal extra cost and achieves state-of-the-art single-model results on the COCO detection benchmark without bells and whistles. Expand
...
1
2
3
4
...