Real-Time Detection of Abnormal Behavior of Escalator Passengers Based on YOLOv5s

Fig. 8. Comparison of detection results of different algorithms. (a) Fast R-CNN detection result; (b) YOLOv3 detection result; (c) YOLOv4 detection result; (d) YOLOv5s detection result; (e) YOLO-STE detection result

Download full size

Fig. 9. Example figure of actual detection results

Download full size

Table 1. Number of labels for each category
View table
Table 1. Number of labels for each category
Class Tarin Test Total
Up 4184 1046 5230
Down 3928 982 4910
Suitcase 2840 710 3550
Stroller 2586 646 3232
Wheelchair 2402 601 3003

Table 2. Ablation experiments

View table

Table 2. Ablation experiments

Model	P /%	R /%	P_{mAP_0.5} /%	N_Params /10⁶	N_FLOPs /G	w /MB
YOLOv5s	93.6	89.7	94.2	7.07	16.5	13.60
ShuffleNetV2	92.4	89.3	93.1	1.55	3.7	3.75
ShuffNetV2 +C3TR	93.8	91.4	94.0	1.64	3.9	3.91
ShuffNetV2 +SE	94.2	90.2	94.8	1.62	3.8	3.85
ShuffNetV2+ C3TR+SE（YOLO-STE）	94.7	93.7	96.1	1.71	3.9	3.97

Table 3. Comparison experiments with common models

View table

Table 3. Comparison experiments with common models

Model	P /%	R /%	P_{mAP_0.5} /%	N_Params /10⁶	N_FLOPs /G	w /MB
Fast R-CNN	74.32	85.62	80.65	22.48	303.6	108.33
YOLOv3	89.87	76.25	88.67	61.55	155.3	234.68
YOLOv4	88.56	79.63	90.62	64.36	134.6	244.53
YOLOv5s	93.60	89.70	94.20	7.07	16.5	13.60
YOLO-STE	94.7	93.70	96.10	1.71	3.9	3.97

Table 4. Specific configuration of Jetson Nano

View table

Table 4. Specific configuration of Jetson Nano

Hardware and software platform	Configuration
Operating system	Ubuntu18.04
CPU	4-core ARM®Cortex®-A57 MPCore
GPU	NVIDIA Maxwell™ with 128 NVIDIA CUDA®core
Graphic memory	4 GB 64 bit LPDDR4
CUDA	10.2
Framework	PyTorch
Programming language	Python3.6

Table 5. Comparison experiments at Jetson Nano
View table
Table 5. Comparison experiments at Jetson Nano
Model Processing /ms Inference /ms NMS /ms $t_{F P}$ /ms
YOLOv5s 2.6 178.5 4.5 185.6
YOLO-STE 2.6 103.2 5.7 111.5

Tools

Get Citation

Copy Citation Text

Yuanpeng Wang, Haibin Wan, Kai Huang, Zhaozhan Chi, Jinqi Zhang, Zhixing Huang. Real-Time Detection of Abnormal Behavior of Escalator Passengers Based on YOLOv5s[J]. Laser & Optoelectronics Progress, 2024, 61(8): 0812004

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites