2020/03/08 by Elad Plaut, Plaut, Elad, Erez Ben Yaacov +3 · 2 citations
Computer Science · Engineering · #Advanced Image and Video Retrieval Techniques #Advanced Neural Network Applications #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #I.4.8 #Robotics and Sensor-Based Localization
paper · pdf · doi:10.48550/arxiv.2003.03759
openalex publication_date 2020/03/08 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Existing monocular 3D object detection methods have been demonstrated on\nrectilinear perspective images and fail in images with alternative projections\nsuch as those acquired by fisheye cameras. Previous works on object detection\nin fisheye images have focused on 2D object detection, partly due to the lack\nof 3D datasets of such images. In this work, we show how to use existing\nmonocular 3D object detection models, trained only on rectilinear images, to\ndetect 3D objects in images from fisheye cameras, without using any fisheye\ntraining data. We outperform the only existing method for monocular 3D object\ndetection in panoramas on a benchmark of synthetic data, despite the fact that\nthe existing method trains on the target non-rectilinear projection whereas we\ntrain only on rectilinear images. We also experiment with an internal dataset\nof real fisheye images.\n