Improved Tiny YOLOv4 Algorithm and Its Application in Pedestrian Detection

Yong Xuan; Chao Han; Wenhan Sha

doi:10.3788/LOP202259.1215002

Laser & Optoelectronics Progress, Volume. 59, Issue 12, 1215002(2022)

Improved Tiny YOLOv4 Algorithm and Its Application in Pedestrian Detection

Yong Xuan¹, Chao Han^1、*, and Wenhan Sha²

¹Key Laboratory of Advanced Perception and Intelligent Control of High-End Equipment, Ministry of Education, Anhui Polytechnic University, Wuhu 241000, Anhui , China

²Chery New Energy Automobile Co., Ltd., Wuhu 241000, Anhui , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(13)

Fig. 1. Network structure of Tiny YOLOv4

Download full size

Fig. 2. Improved network structure

Download full size

Fig. 3. Network structures of traditional convolution and depthwise separable convolution. (a) Network structure of traditional convolution; (b) network structure of depthwise separable convolution

Download full size

Fig. 4. Channel attention structure

Download full size

Fig. 5. Structure diagram of spatial attention module

Download full size

Fig. 6. Feature enhancement module

Download full size

Fig. 7. Loss values of different network structures. (a) Loss values of YOLOv4; (b) loss values of Tiny YOLOv4; (c) loss values of improved network

Download full size

Fig. 8. Experimental results

Download full size

Table 1. Comparison of training parameters of different network models
View table
Table 1. Comparison of training parameters of different network models
Detection algorithm Number of training parameters
YOLOv4 64040001
Tiny YOLOv4 5939804
Proposed Tiny YOLOv4 4369606

Table 2. Size comparison of different network models
View table
Table 2. Size comparison of different network models
Detection algorithm Model size /MB
YOLOv4 246
Tiny YOLOv4 22.7
Proposed Tiny YOLOv4 15.9

Table 3. Test results of different algorithms on different datasets

View table

Table 3. Test results of different algorithms on different datasets

Dataset	Algorithm	AP /%	FPS /（frame·s^-1）
INRIA	YOLOv4	86.2	20.1
	Tiny YOLOv4	68.67	31.6
	Proposed Tiny YOLOv4	76.32	30.4
COCO	YOLOv4	90.76	20.8
	Tiny YOLOv4	69.76	32.4
	Proposed Tiny YOLOv4	78.2	30.2
VOC	YOLOv4	90.4	19.8
	Tiny YOLOv4	71.27	31.2
	Proposed Tiny YOLOv4	78.8	30.6
Mixed data	YOLOv4	92.5	19.5
	Tiny YOLOv4	73.42	32.2
	Proposed Tiny YOLOv4	80.52	31.4

Table 4. Comparison of results of detection algorithm

View table

Table 4. Comparison of results of detection algorithm

Algorithm	AP /%	Recall /%	FPS /（frame·s^-1）
Faster R-CNN	72.37	78.4	4.7
SSD	78.20	79.1	22.1
YOLOv3	85.34	77.6	17.3
Tiny YOLOv3	67.80	73.4	25.6
YOLOv4	92.50	81.5	19.5
Tiny YOLOv4	73.42	75.7	32.2
Proposed Tiny YOLOv4	80.52	82.3	31.4

Table 5. Ablation experiments on mixed datasets
View table
Table 5. Ablation experiments on mixed datasets
Tiny YOLOv4 baseline DSC Attention mechanism Scale enhancement AP /%
√ 73.42
√ √ 74.32
√ √ √ 77.27
√ √ √ √ 80.52

Tools

Get Citation

Copy Citation Text

Yong Xuan, Chao Han, Wenhan Sha. Improved Tiny YOLOv4 Algorithm and Its Application in Pedestrian Detection[J]. Laser & Optoelectronics Progress, 2022, 59(12): 1215002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites