Abstract
We propose a novel method for detecting horizontal van-ishing points and the zenith vanishing point in man-madeenvironments. The dominant trend in existing methods isto first find candidate vanishing points, then remove out-liers by enforcing mutual orthogonality. Our method re-verses this process: we propose a set of horizon line can-didates and score each based on the vanishing points it contains. A key element of our approach is the use of global image context, extracted with a deep convolutional network, to constrain the set of candidates under consideration. Our method does not make a Manhattan-world assumption and can operate effectively on scenes with only a single horizontal vanishing point. We evaluate our approach on three benchmark datasets and achieve state-of-the-art performance on each. In addition, our approach issignificantly faster than the previous best method.