Object Detection in Equirectangular Panorama

We introduce a high-resolution equirectangular panorama (aka 360-degree, virtual reality, VR) dataset for object detection and propose a multi-projection variant of the YOLO detector. The main challenges with equirectangular panorama images are i) the lack of annotated training data, ii) high-resolu...

Full description

Saved in:
Bibliographic Details
Published in2018 24th International Conference on Pattern Recognition (ICPR) pp. 2190 - 2195
Main Authors Yang, Wenyan, Qian, Yanlin, Kamarainen, Joni-Kristian, Cricri, Francesco, Fan, Lixin
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.08.2018
Subjects
Cameras
Detectors
Distortion
Graphics processing units
Image resolution
Object detection
Virtual reality
Online AccessGet full text
DOI10.1109/ICPR.2018.8546070

Cover

More Information
Summary:We introduce a high-resolution equirectangular panorama (aka 360-degree, virtual reality, VR) dataset for object detection and propose a multi-projection variant of the YOLO detector. The main challenges with equirectangular panorama images are i) the lack of annotated training data, ii) high-resolution imagery and iii) severe geometric distortions of objects near the panorama projection poles. In this work, we solve the challenges by I) using training examples available in the "conventional datasets" (ImageNet and COCO), II) employing only low resolution images that require only moderate GPU computing power and memory, and III) our multi-projection YOLO handles projection distortions by making multiple stereographic sub-projections. In our experiments, YOLO outperforms the other state-of-the-art detector, Faster R-CNN, and our multi-projection YOLO achieves the best accuracy with low-resolution input.
DOI:10.1109/ICPR.2018.8546070