Scholarlink

To understand the overexposure effect of road segmentation efficiency when self-driving sweeper bot (SDSB) runs at a sunshine light scenery, this work collected the normal exposed image and the overexposed image from SDSB to investigate the trade-off of training cost on two types of datasets and set up epoch from 140 to 700 step 140 for validation on three extension U-net methods. The experimental results indicated that in the overexposed case. However, it involved a higher epoch of training cost, and the segmented efficiency is still worse than the normal exposed one. This further implied the challenge of SDSB vision sensors to visualize controlling the SDSB running on the constrained road in sunshine status.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing Training Epoch on Overexposed Road Segmentation of Fish-Eye Visual Sensors

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Xiaodong Yu,
  • Ta-Wen Kuan,
  • Ying Chen,
  • Jhing-Fa Wang,
  • Tuoli Chen,
  • Qilong Huang

摘要

To understand the overexposure effect of road segmentation efficiency when self-driving sweeper bot (SDSB) runs at a sunshine light scenery, this work collected the normal exposed image and the overexposed image from SDSB to investigate the trade-off of training cost on two types of datasets and set up epoch from 140 to 700 step 140 for validation on three extension U-net methods. The experimental results indicated that in the overexposed case. However, it involved a higher epoch of training cost, and the segmented efficiency is still worse than the normal exposed one. This further implied the challenge of SDSB vision sensors to visualize controlling the SDSB running on the constrained road in sunshine status.