TY  - JOUR
T1  - Efficient Object Detection Models for Autonomous Driving System
AU  - Kim, Si-On 
AU  - Lee, Sun-Young 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2025
DA  - 2025/1/1
DO  - 10.7840/kics.2025.50.1.149
KW  - autonomous Driving System
KW  - Object Encoding
KW  - YOLOv5
KW  - 100Base-T1
AB  - The development of autonomous driving technology has also been accelerated by object detection using big data processing and machine learning. Numerous sensors and ECUs were installed in the vehicle for autonomous driving, and cables for communication between each sensor and ECU were installed, causing problems such as increasing the weight of the vehicle and reducing fuel efficiency. CAN, the longest-used vehicle network, has the disadvantage that it is not suitable for real-time video transmission with a maximum transmission speed of 1 Mbps. The Ethernet for vehicles used to solve this problem can transmit image data at a maximum transmission speed of 100 Mbps, but it has the disadvantage that it is difficult to transmit high-resolution images. In this study, after deleting non-detection object data, a method of reducing data transmission time while maintaining object detection performance by object encoding the area data of the detection object was proposed and evaluated through experiments. As a result, it was confirmed that the transmission time was reduced by 41.02% in the FHD environment and 62.8% in the 4K environment.