SLNet: A Semantic Line Detector with
Multi-Task and Multi-Scale Learning

Jun-Tae Lee
jtlee@mcl.korea.ac.kr
Korea University

Han-Ul Kim
hanulkim@mcl.korea.ac.kr
Korea University

Chul Lee
chullee@pknu.ac.kr
Pukyong National University

Chang-Su Kim
changsukim@korea.ac.kr
Korea University

Abstract

Semantic lines characterize the layout of an image. Despite their importance in image analysis and scene understanding, there is no reliable research for semantic line detection. In this paper, we propose a semantic line detector using a convolutional neural network with multi-task learning, by regarding the line detection as a combination of classification and regression tasks. We use convolution and max-pooling layers to obtain multi-scale feature maps for an input image. Then, we develop the line pooling layer to extract a feature vector for each candidate line from the feature maps. Next, we feed the feature vector into the parallel classification and regression layers. The classification layer decides whether the line candidate is semantic or not. In case of a semantic line, the regression layer determines the offset for refining the line location. Experimental results show that the proposed detector extracts semantic lines accurately and reliably. Moreover, we demonstrate that the proposed detector can be used successfully in three applications: horizon estimation, composition enhancement, and image simplification.

Supplementary Video

                             

Publications

Jun-Tae Lee, Han-Ul Kim, Chul Lee, and Chang-Su Kim, "Semantic Line Detection and Its Applications," in Proc. International Conference on Computer Vision (ICCV), Venice, Italy, pp. 3229-3237, Oct. 2017. [code]

Jun-Tae Lee, Han-Ul Kim, Chul Lee, and Chang-Su Kim, "SLNet: A Semantic Line Detector with Multi-Task and Multi-Scale Learning," submitted to IEEE Access.

Dataset

You may download the SEL dataset from here.