Abstract:With the development of self-driving technology, pedestrian detection models are becoming more useful and important. Although object detection models based on Convolutional Neural Networks have achieved favorable detection results, due to the occlusion in crowded scenarios, dense pedestrian detection is still a challenging task compared with generic object detection. For a CNN-based object detector, an appropriate label assignment method is significant. In this paper, we propose an effective training sample allocation method called Occlusion-Prediction aware Label Assignment (OPLA) to improve the performance of the detector in crowded scenarios. For those Ground Truths (GTs) without severe occlusion, we directly assign prediction boxes with the best matching quality as the corresponding positive samples. In terms of GTs in heavily crowded areas, we fine-tune the matching quality based on visible ratio to enhance the training sample selection priority of severe occluded GTs. In addition, we reveal the deficiency of currently popular post-processing algorithms, Non-Maximum Suppression (NMS) and Soft-NMS for dense pedestrian detection, and propose Hierarchical-NMS (H-NMS) whose strategy diverges with different Intersection over Union (IoU) levels. Experimental results on Crowdhuman, the largest dense pedestrian detection dataset at present, show our approach optimizes the Average Precision (AP) and Miss Rate (MR) of Fully Convolutional One Stage Object Detection (FCOS) by 4.05% and 5.35% and outperforms many state-of-the-art detectors, respectively. Besides, our method does not bring any increase in calculation during inference.

An Enhanced Histogram of Oriented Gradients for Pedestrian Detection

Pedestrian detection using improved Histogram of Oriented Gradients

A Method for Vehicle Detection Based on Local Gradients Vector

Research on Hierarchical Pedestrian Detection Based on SVM Classifier with Improved Kernel Function

Pedestrian detection based on modified dynamic background using gaussian mixture models and hog-svm detection

A New Pedestrian Detection Method Based on Combined HOG and LSS Features

Towards Accurate Dense Pedestrian Detection Via Occlusion-Prediction Aware Label Assignment and Hierarchical-Nms.

A new depth descriptor for pedestrian detection in RGB-D images

Pedestrian Detection Method Based on Improved YOLOv5s for Densely Occluded Scenarios

Fast Pedestrian Detection And Tracking Based On Vibe Combined Hog-Svm Scheme

An HOG-CT Human Detector with Histogram-Based Search.

Effective Pedestrian Detection Using Center-symmetric Local Binary/Trinary Patterns

Steerable Second Order Intensity Features For Pedestrian Detection

Pedestrian Detection in Complex Scene Using Full Binary Tree Classifiers Based on Locally Assembled Binary Haar-Like Features

Pedestrian Detection Via Part-Based Topology Model

Color Self-Similarity Feature Based Real-Time Pedestrian Detection

Pedestrian Detection Optimization Algorithm Based on Low-Altitude UAV

Pedestrian Detection Based on Hybrid Features

A Pedestrian-Detection Method Based on Heterogeneous Features and Ensemble of Multi-View–Pose Parts

An Experimental Evaluation of Local Features for Pedestrian Classification

Pyramid Center-Symmetric Local Binary/Trinary Patterns for Effective Pedestrian Detection