Regression with residual neural network for vanishing point detection

摘要

This paper aims to propose a regression method with a residual neural network (ResNet) for vanishing point detection. The purpose of this study is to estimate the position of the vanishing point accurately. Our newly collected Naver Maps' Street View dataset is used for training regression ResNet-34 and for comparison with previous methods. It is concluded that the trained regression ResNet outperforms previous methods in terms of both computation time and accuracy.