Lightweight convolutional neural network-based pedestrian detection and re-identification in multiple scenarios

Xiao, Ke; Lin, Xiao; Qin, Long

doi:10.1007/s00138-021-01169-7

Cited by 13 publications

(6 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The backbone uses MobileNetv2, which employs a channel separation aggregation module that can simplify stacking and separable convolution complexity and a dynamic receptive field module that can improve the performance of feature extraction by dynamically customizing the convolution core and its perception range. The authors of [29] proposed a pedestrian detection method based on data augmentation that combines a lightweight CNN (LCNN)-based pedestrian recognition model with YOLOv3 to enlarge the image size of pedestrians smaller than the specified value. This end-to-end framework improves detection stability and accuracy.…”

Section: Hybrid Approachesmentioning

confidence: 99%

Object Detection Network Based on Module Stack and Attention Mechanism

Dou,

Wang,

Shao

et al. 2023

Electronics

View full text Add to dashboard Cite

Currently, visual computer applications based on convolutional neural networks are rapidly developing. However, several problems remain: (1) high-quality graphics processing equipment is needed, and (2) the trained network model has several unnecessary convolution operations. These problems result in a single-stage target detection network that often requires unnecessary computing power and is difficult to apply to equipment with insufficient computing resources. To solve these problems, based on YOLOv5, a YOLOv5-L (YOLOv5 Lightweight) network structure is proposed. This network is improved using YOLOv5. First, to enhance the inference speed of the detector on the CPU, the PP-LCNet (PaddlePaddle-Lightweight CPU Net) is employed as the backbone network. Second, the focus module is removed, and the end convolution module in the head network is replaced by a deep separable convolution module, which eliminates redundant operations and reduces the amount of computation. The experimental results show that YOLOv5-L enables a 48% reduction in model parameters and computation compared to YOLOv5, a 35% increase in operation speed, and a less than 2% reduction in accuracy, which is significant in the environment of low-performance computing equipment.

show abstract

Section: Hybrid Approachesmentioning

confidence: 99%

Object Detection Network Based on Module Stack and Attention Mechanism

Dou,

Wang,

Shao

et al. 2023

Electronics

View full text Add to dashboard Cite

show abstract

“…It has obtained a fully convolutional pedestrian detection model that can be run on low computational resources. Ke [22] processed with a nonoverlapped image blocking data augmentation method, and then input them into the YOLOv3 detector to obtain the object position information. An LCNN-based pedestrian re-identification model is used to extract the features of the object.…”

Section: Research On Lightweight Target Detection Algorithmsmentioning

confidence: 99%

SMYOLO: Lightweight Pedestrian Target Detection Algorithm in Low-Altitude Scenarios

Zhang

et al. 2022

Future Internet

View full text Add to dashboard Cite

Due to the arbitrariness of the drone’s shooting angle of view and camera movement and the limited computing power of the drone platform, pedestrian detection in the drone scene poses a greater challenge. This paper proposes a new convolutional neural network structure, SMYOLO, which achieves the balance of accuracy and speed from three aspects: (1) By combining deep separable convolution and point convolution and replacing the activation function, the calculation amount and parameters of the original network are reduced; (2) by adding a batch normalization (BN) layer, SMYOLO accelerates the convergence and improves the generalization ability; and (3) through scale matching, reduces the feature loss of the original network. Compared with the original network model, SMYOLO reduces the accuracy of the model by only 4.36%, the model size is reduced by 76.90%, the inference speed is increased by 43.29%, and the detection target is accelerated by 33.33%, achieving minimization of the network model volume while ensuring the detection accuracy of the model.

show abstract

“…However, with the complexity and diversity of human body posture, the problems of insufficient light and occlusion are more serious. Hence, it is of great difficulty to accurately detect pedestrians in various scenes [2]. Pedestrians are constantly moving in video frames or images, resulting in different positions and attitudes of the same pedestrian in different video frames or images.…”

Section: Introductionmentioning

confidence: 99%

“…(1) Local occlusion will greatly reduce the amount of information required for detection, leading to missed detection [7]. (2) It is difficult to extract features with strong discrimination from small-size pedestrians, leading to unsatisfactory detection results. Presently, most of the frontier pedestrian detection research work is based on deep learning.…”

Section: Introductionmentioning

confidence: 99%

Pedestrian Detection Algorithm Combining Attention Mechanism and Nonmaximum Suppression Method

Pan¹,

Zhou²

2022

Advances in Multimedia

View full text Add to dashboard Cite

With the involution of pedestrian detection technology, higher requirements are put forward for the detection accuracy under the conditions of insufficient light, target occlusion, and too small scale. Without information and multiscale pedestrian target, visible light single-mode pedestrian detection algorithm has poor performance. To solve the above problems, a pedestrian detection algorithm combining attention mechanism and nonmaximum suppression method is proposed in this study, aiming to improve the accuracy of pedestrian detection. In addition, residual network ResNet-50 and IoU (intersection over union) loss function are also adopted to improve pedestrian detection accuracy. Attention mechanism was used to optimize and highlight pedestrian area features, and meanwhile, the nonmaximum suppression method was applied to improve the robustness of the algorithm. Experimental results show that the detection accuracy of the proposed algorithm is significantly higher than that of the traditional convolutional neural network algorithm.

show abstract

Lightweight convolutional neural network-based pedestrian detection and re-identification in multiple scenarios

Cited by 13 publications

References 25 publications

Object Detection Network Based on Module Stack and Attention Mechanism

Object Detection Network Based on Module Stack and Attention Mechanism

SMYOLO: Lightweight Pedestrian Target Detection Algorithm in Low-Altitude Scenarios

Pedestrian Detection Algorithm Combining Attention Mechanism and Nonmaximum Suppression Method

Contact Info

Product

Resources

About