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.

Semi-supervised Elastic net for pedestrian counting

A Double-Region Learning Algorithm for Counting the Number of Pedestrians in Subway Surveillance Videos

Semisupervised Pedestrian Counting With Temporal and Spatial Consistencies

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

Improved SENet for Pedestrian Detection

Pedestrian detection with unsupervised multispectral feature learning using deep neural networks

Dynamic Region Division for Adaptive Learning Pedestrian Counting

Predicting Pedestrian Counts in Crowded Scenes With Rich and High-Dimensional Features

Dual-branch Self-Attention Network for Pedestrian Attribute Recognition

Counting Pedestrian with Mixed Features and Extreme Learning Machine

Count- and Similarity-Aware R-CNN for Pedestrian Detection

GoogLeNet-like Model for Pedestrian Attribute Detection in Surveillance Environment

A Pedestrian Classification Method Based on Transfer Learning

Pedestrian Counting System Based On Multiple Object Detection And Tracking

Weakly Supervised Pedestrian Attribute Recognition with Attention in Latent Space

Generate and adjust: a novel framework for semi-supervised pedestrian attribute recognition

Transfer Learning Based on Sparse Coding and It's Application in Pedestrian Detection

Pedestrian Detection Based on Sparse Coding and Transfer Learning

Discriminative Latent Semantic Feature Learning for Pedestrian Detection

A pedestrian detection algorithm based on deep learning

CSGAT-Net: a conditional pedestrian trajectory prediction network based on scene semantic maps and spatiotemporal graph attention