SS-SFR:用于目标检测的虚拟 KITTI 上合成场景空间频率响应
计算机视觉与模式识别
2024-10-02 v2
摘要
汽车模拟可以potentially弥补计算机视觉应用中训练数据不足的问题。然而,针对汽车模拟的图像质量评估仍很少,而光学退化对模拟的影响亦未得到充分探索。本文研究了 Virtual KITTI 以及应用高斯模糊变化对图像锐度的影响。此外,我们考虑目标检测,这是一种常见的计算机视觉应用,涉及三个最新模型,从而使我们能够描述目标检测与锐度之间的关系。发现虽然图像锐度(MTF50)从平均 0.245cy/px 下降到约 0.119cy/px,但目标检测性能在所有相应保留测试集中基本保持稳健,分别为 Faster RCNN(0.58%)、YOLOF(1.45%)和 DETR(1.93%)。
引用
@article{arxiv.2407.15646,
title = {SS-SFR: Synthetic Scenes Spatial Frequency Response on Virtual KITTI and Degraded Automotive Simulations for Object Detection},
author = {Daniel Jakab and Alexander Braun and Cathaoir Agnew and Reenu Mohandas and Brian Michael Deegan and Dara Molloy and Enda Ward and Tony Scanlan and Ciarán Eising},
journal= {arXiv preprint arXiv:2407.15646},
year = {2024}
}
备注
8 pages, 2 figures, 2 tables. This paper is a preprint of a paper submitted to the 26th Irish Machine Vision and Image Processing Conference (IMVIP 2024). If accepted, the copy of record will be available at IET Digital Library