Have a personal or library account? Click to login
Road Obstacle Object Detection Based on Improved YOLO V4 Cover
By: Xiao Zuo,  Jun Yu,  Tong Xian,  Yuzhe Hu and  Zhiyi Hu  
Open Access
|Feb 2021

Figures & Tables

Figure 1.

Network structure of YOLO v4 algorithm
Network structure of YOLO v4 algorithm

Figure 2.

Experimental framework
Experimental framework

Figure 3.

Annotation of car instance segmentation
Annotation of car instance segmentation

Figure 4.

The json file of the car label
The json file of the car label

Figure 5.

The txt file of the image tag
The txt file of the image tag

Figure 6.

Add Gaussian noise
Add Gaussian noise

Figure 7.

Median fuzzy processing
Median fuzzy processing

Figure 8.

Object number before and after data amplification
Object number before and after data amplification

Figure 9.

Learning rate change curve
Learning rate change curve

Figure 10.

Experimental results based on Yolo v4
Experimental results based on Yolo v4

Figure 11.

Experimental results based on improved YOLO v4
Experimental results based on improved YOLO v4

Comparison of test results

AP(%)CarBusPersonMotorbikeBicyclemAP(%)
yolo v40.980.930.920.810.5182.95
Improved yolo v40.990.930.920.810.5884.98

Environment configuration

Hardware environmentprocessorGraphics cardIntel(R) XEON W-2133Nvidia TITAN XP 12G
Software Environmentoperating systemUbuntu 16.04
Deep learning frameworkTensorflow-gpu
Programming languagePython
translaterPycharm2019.1

Main network parameter values

ParameterValueParameterValue
LEARN_RATE_INIT1e-4MOVING_AVE_DECAY0.9999
LEARN_RATE_END1e-6STAGE_EPOCHS100
Language: English
Page range: 18 - 25
Published on: Feb 22, 2021
In partnership with: Paradigm Publishing Services
Publication frequency: 4 issues per year

© 2021 Xiao Zuo, Jun Yu, Tong Xian, Yuzhe Hu, Zhiyi Hu, published by Xi’an Technological University
This work is licensed under the Creative Commons Attribution 4.0 License.