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

arXiv: https://arxiv.org/abs/2106.02426 · HTML: https://mlsanity.com/paper/2106.02426/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 37 references that are in the map.
- First posted on arXiv 4 Jun 2021.

## Builds on (most cited first, 22 of 22)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://mlsanity.com/paper/1409.1556/) (2014)
- [Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks](https://mlsanity.com/paper/1506.01497/) (2015)
- [Soft-NMS -- Improving Object Detection With One Line of Code](https://mlsanity.com/paper/1704.04503/) (2017)
- [CityPersons: A Diverse Dataset for Pedestrian Detection](https://mlsanity.com/paper/1702.05693/) (2017)
- [A Unified Multi-scale Deep Convolutional Neural Network for Fast Object Detection](https://mlsanity.com/paper/1607.07155/) (2016)
- [Adaptive NMS: Refining Pedestrian Detection in a Crowd](https://mlsanity.com/paper/1904.03629/) (2019)
- [Detection in Crowded Scenes: One Proposal, Multiple Predictions](https://mlsanity.com/paper/2003.09163/) (2020)
- [Occlusion-aware R-CNN: Detecting Pedestrians in a Crowd](https://mlsanity.com/paper/1807.08407/) (2018)
- [Is Faster R-CNN Doing Well for Pedestrian Detection?](https://mlsanity.com/paper/1607.07032/) (2016)
- [Mask-Guided Attention Network for Occluded Pedestrian Detection](https://mlsanity.com/paper/1910.06160/) (2019)
- [NMS by Representative Region: Towards Crowded Pedestrian Detection by Proposal Pairing](https://mlsanity.com/paper/2003.12729/) (2020)
- [PedHunter: Occlusion Robust Pedestrian Detector in Crowded Scenes](https://mlsanity.com/paper/1909.06826/) (2019)
- [Accurate Single Stage Detector Using Recurrent Rolling Convolution](https://mlsanity.com/paper/1704.05776/) (2017)
- [What Can Help Pedestrian Detection?](https://mlsanity.com/paper/1705.02757/) (2017)
- [Illuminating Pedestrians via Simultaneous Detection & Segmentation](https://mlsanity.com/paper/1706.08564/) (2017)
- [Relational Learning for Joint Head and Human Detection](https://mlsanity.com/paper/1909.10674/) (2019)
- [Pedestrian Detection with Autoregressive Network Phases](https://mlsanity.com/paper/1812.00440/) (2018)
- [Filtered Channel Features for Pedestrian Detection](https://mlsanity.com/paper/1501.05759/) (2015)
- [Fused DNN: A deep neural network fusion approach to fast and robust pedestrian detection](https://mlsanity.com/paper/1610.03466/) (2016)
- [SSA-CNN: Semantic Self-Attention CNN for Pedestrian Detection](https://mlsanity.com/paper/1902.09080/) (2019)
- [Where, What, Whether: Multi-modal Learning Meets Pedestrian Detection](https://mlsanity.com/paper/2012.10880/) (2020)

## Built on by (most cited first, 7 of 7)

- [Beyond Appearance: a Semantic Controllable Self-Supervised Learning Framework for Human-Centric Visual Tasks](https://mlsanity.com/paper/2303.17602/) (2023)
- [VLPD: Context-Aware Pedestrian Detection via Vision-Language Semantic Self-Supervision](https://mlsanity.com/paper/2304.03135/) (2023)
- [FusionVision: A comprehensive approach of 3D object reconstruction and segmentation from RGB-D cameras using YOLO and fast segment anything](https://mlsanity.com/paper/2403.00175/) (2024)
- [Revisiting the Evaluation of Deep Neural Networks for Pedestrian Detection](https://mlsanity.com/paper/2511.10308/) (2025)
- [Research on road object detection algorithm based on improved YOLOX](https://mlsanity.com/paper/2302.08156/) (2023)
- [From 2D to 3D Without Extra Baggage: Data-Efficient Cancer Detection in Digital Breast Tomosynthesis](https://mlsanity.com/paper/2511.10597/) (2025)
- [A Probabilistic Framework for Improving Dense Object Detection in Underwater Image Data via Annealing-Based Data Augmentation](https://mlsanity.com/paper/2604.21198/) (2026)

## Abstract (arXiv)

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.