Road Object Detection Robust to Distorted Objects at the Edge Regions of Images

Wooksu Shin, Donghyuk Choi, Hancheol Park, Jeongho Kim; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2024, pp. 7245-7251

Fish-eye cameras capable of capturing wide areas en- able efficient traffic monitoring with only a few cameras. Nevertheless it still remains challenging to successfully de- tect objects in images from such cameras. In this work we analyze key reasons why object detectors frequently make incorrect predictions in such images and propose methods to address them. More specifically we address the issues of objects being represented as smaller at the edges of images and the distortion of non-target objects (e.g. street signs) which are recognized as target objects (e.g. vehicles). Fur- thermore in this work we propose a road object detector capable of achieving high performance by additionally ap- plying various techniques known to generally enhance de- tection performance. Our proposed detector achieved sec- ond place in Track 4 of the 2024 AI City Challenge with an F1 score of 0.6196. Our code is publicly available at https://github.com/nota-github/AIC2024_ Track4_Nota.