Unbiased IoU for Spherical Image Object Detection

As one of the fundamental components object detection, intersection-over-union (IoU) calculations between two bounding boxes play an important role in samples selection, NMS operation and evaluation detection algorithms. This procedure is well-defined solved for planar images, while it challenging spherical ones. Some existing methods utilize to represent objects. However, they are biased due distortions Others use rectangles as unbiased representations, but adopt excessive approximate algorithms when computing IoU. In this paper, we propose IoU a novel criterion image which based on representations analytical method calculation. first time that absolutely accurate calculation applied criterion, thus can be correctly evaluated images. With representation calculation, also present Spherical CenterNet, anchor free algorithm The experiments show our gives results proposed CenterNet achieves better performance real-world synthetic datasets than methods.