NMS by Representative Region: Towards Crowded Pedestrian Detection by Proposal Pairing

Huang, Xin; Ge, Zheng; Jie, Zequn; Yoshie, Osamu

doi:10.1109/cvpr42600.2020.01076

Cited by 154 publications

(86 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…OR-CNN [39] proposes a part-aware RoI pooling unit to integrate the prior structure information of the human body with visibility prediction into the Fast R-CNN module of the detector. Xie et al [64] [60] ResNet-50 CH/COCOPersons Two-stage Anchor-Based arXiv2019 JointDet [61] ResNet-50 CP/CH/CA Two-stage Anchor-Based AAAI2020 PedHunter [25] ResNet-50 CA/CP/CH Two-stage Anchor-Based AAAI2020 PRNet [62] ResNet-50 CA/CP/ETH Single-stage Anchor-Based ECCV2020 FC-Net [63] ResNet-50 CA/CP Two-stage Anchor-Based ITS2020 PSC [64] VGG16 CA/CP Two-stage Anchor-Based arXiv2020 V2F-Net [65] ResNet-50 [74] Resnet-50/VGG16 CP/CH Two-stage Anchor-Based CVPR2020 NOH NMS [75] ResNet-50 CP/CH -Anchor-Based ACM MM2020 APD [76] ResNet-50/DLA34 CP/CH Single-stage Anchor-Free TMM 2020 MAPD [77] ResNet-50 CP/CH Single-stage Anchor-Free Neurocomputing LLA [78] ResNet-50/ResNet-101 CP/CH -Based/Free arXiv2021 NMS-ped [79] Resnet-50 CA/CP -Anchor-Based arXiv2021…”

Section: A Part-based Methodsmentioning

confidence: 99%

Occlusion Handling and Multi-Scale Pedestrian Detection Based on Deep Learning: A Review

Fang

Liu

et al. 2022

IEEE Access

View full text Add to dashboard Cite

Pedestrian detection is an important branch of computer vision, and it has important applications in the fields of autonomous driving, artificial intelligence and video surveillance. With the rapid development of deep learning and the proposal of large-scale datasets, pedestrian detection has reached a new stage and achieves better performance. However, the performance of state-of-the-art methods is far behind the expectation, especially when occlusion and scale variance exist. Therefore, a lot of works focused on occlusion and scale variance have been proposed in the past few years. The purpose of this article is to make a detailed review of recent progress in pedestrian detection. Firstly, brief progress of pedestrian detection in the past two decades is summarized. Secondly, recent deep learning methods focusing on occlusion and scale variance are analyzed. Moreover, the popular datasets and evaluation methods for pedestrian detection are introduced. Finally, the development trend of pedestrian detection is prospected.

show abstract

Section: A Part-based Methodsmentioning

confidence: 99%

Occlusion Handling and Multi-Scale Pedestrian Detection Based on Deep Learning: A Review

Fang

Liu

et al. 2022

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Some methods [58], [8], [52], [66] aim to improve small-scale pedestrian detection, while some methods [78], [83], [51] exploit the part or visible information for occluded pedestrian detection. To improve pedestrian detection in crowded scenes, some methods [63], [45], [33], [11] exploit how to combine the highly overlapping bounding boxes.…”

Section: B the Methods Of Object Detectionmentioning

confidence: 99%

TJU-DHD: A Diverse High-Resolution Dataset for Object Detection

Pang

Cao

et al. 2021

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

Vehicles, pedestrians, and riders are the most important and interesting objects for the perception modules of selfdriving vehicles and video surveillance. However, the state-of-theart performance of detecting such important objects (esp. small objects) is far from satisfying the demand of practical systems. Large-scale, rich-diversity, and high-resolution datasets play an important role in developing better object detection methods to satisfy the demand. Existing public large-scale datasets such as MS COCO collected from websites do not focus on the specific scenarios. Moreover, the popular datasets (e.g., KITTI and Citypersons) collected from the specific scenarios are limited in the number of images and instances, the resolution, and the diversity. To attempt to solve the problem, we build a diverse high-resolution dataset (called TJU-DHD). The dataset contains 115,354 high-resolution images (52% images have a resolution of 1624×1200 pixels and 48% images have a resolution of at least 2,560×1,440 pixels) and 709,330 labeled objects in total with a large variance in scale and appearance. Meanwhile, the dataset has a rich diversity in season variance, illumination variance, and weather variance. In addition, a new diverse pedestrian dataset is further built. With the four different detectors (i.e., the one-stage RetinaNet, anchor-free FCOS, two-stage FPN, and Cascade R-CNN), experiments about object detection and pedestrian detection are conducted. We hope that the newly built dataset can help promote the research on object detection and pedestrian detection in these two scenes. The dataset is available at https://github.com/tjubiit/TJU-DHD.

show abstract

“…Backbone MR RepLoss [31] ResNet-50 13.20% OR-CNN [38] ResNet-50 12.80% Adaptive-NMS [19] VGG-16 11.90% CSP [21] ResNet-50 11.00% MGAN [25] VGG-16 11.50% R 2 NMS [18] VGG-16 11.10% EMD-RCNN [7] ResNet-50 10.70% Our baseline…”

Section: Methodsmentioning

confidence: 99%

“…The second one uses deep convolutional neural networks (CNNs) to obtain highlevel semantic feature representation, which has a discriminative ability to deal with complex scenes for pedestrian detection. To alleviate FN issue in high occlusion scenes, different variants of Non-Maximum Suppression (NMS) [1,18,19] are proposed to change NMS threshold during inference adaptively. To reduce FP, many works [5,6] jointly predict pedestrian boxes and parts information such as head due to that it is less occluded.…”

Section: Introductionmentioning

confidence: 99%

NMS-Loss: Learning with Non-Maximum Suppression for Crowded Pedestrian Detection

Luo

Fang

Zheng

et al. 2021

Proceedings of the 2021 International Conference on Multimedia Retrieval

View full text Add to dashboard Cite

Non-Maximum Suppression (NMS) is essential for object detection and affects the evaluation results by incorporating False Positives (FP) and False Negatives (FN), especially in crowd occlusion scenes. In this paper, we raise the problem of weak connection between the training targets and the evaluation metrics caused by NMS and propose a novel NMS-Loss making the NMS procedure can be trained end-to-end without any additional network parameters. Our NMS-Loss punishes two cases when FP is not suppressed and FN is wrongly eliminated by NMS. Specifically, we propose a pull loss to pull predictions with the same target close to each other, and a push loss to push predictions with different targets away from each other. Experimental results show that with the help of NMS-Loss, our detector, namely NMS-Ped, achieves impressive results with Miss Rate of 5.92% on Caltech dataset and 10.08% on CityPersons dataset, which are both better than state-of-the-art competitors. CCS CONCEPTS• Computing methodologies → Object detection.

show abstract

NMS by Representative Region: Towards Crowded Pedestrian Detection by Proposal Pairing

Cited by 154 publications

References 22 publications

Occlusion Handling and Multi-Scale Pedestrian Detection Based on Deep Learning: A Review

Occlusion Handling and Multi-Scale Pedestrian Detection Based on Deep Learning: A Review

TJU-DHD: A Diverse High-Resolution Dataset for Object Detection

NMS-Loss: Learning with Non-Maximum Suppression for Crowded Pedestrian Detection

Contact Info

Product

Resources

About