Vanishing point detection with convolutional neural networks

TL;DR

This paper presents a convolutional neural network approach for detecting vanishing points in natural scenes, demonstrating improved accuracy over traditional methods by training on a large dataset of annotated road images.

Contribution

The paper introduces an end-to-end CNN model for vanishing point detection in natural environments, advancing beyond classic detection techniques.

Findings

CNN approach outperforms traditional methods

Effective on large-scale YouTube road image dataset

Shows potential for driver assistance applications

Abstract

Inspired by the finding that vanishing point (road tangent) guides driver's gaze, in our previous work we showed that vanishing point attracts gaze during free viewing of natural scenes as well as in visual search (Borji et al., Journal of Vision 2016). We have also introduced improved saliency models using vanishing point detectors (Feng et al., WACV 2016). Here, we aim to predict vanishing points in naturalistic environments by training convolutional neural networks in an end-to-end manner over a large set of road images downloaded from Youtube with vanishing points annotated. Results demonstrate effectiveness of our approach compared to classic approaches of vanishing point detection in the literature.