Training Real-Time Panoramic Object Detectors with Virtual Dataset

With the rapid development of autonomous driving, real-time object detection on 360° images becomes more and more important. In this paper, we propose a panoramic virtual dataset for training object detectors on 360° images. The most important feature of our dataset includes (1) an auto-generated city scene is created for rendering 360° dataset. (2) annotation work for this dataset is automatic. In addition, we propose a modified YOLOv3 model called Pano-YOLO for real-time panoramic object detection. Compared with YOLOv3, mAP of Pano-YOLO drops 0.39%. While speed is 32.47% faster. Experiments are performed to show that models trained on our virtual dataset can be applied in real world. And Pano-YOLO is capable of real-time object detection task on high-resolution 360° panoramic images and videos.

1. INTRODUCTION

In recent years, the number of 360° panoramic images has increased rapidly. With deep learning methods playing a more and more important role in autonomous driving. Object detectors on 360° panoramic images can provide vehicles with full sense of surrounding environment without any dead spot. Equirectangular projection (ERP) is the most used panorama format. However, its easy projection results in distortions around the polar region of images as shown in Figure 1, creating challenges for object detection on 360° images.