ContourGAN: Image contour detection with generative adversarial network

作者：

Highlights：

• We add the adversarial model to image contour detection.

• We employ all convolutional layers in the encoder stage to improve convergence rates.

• We randomly crop 224 × 224 regions and then rescale them to improve fitting ability.

• We evaluated the weight of adversarial loss showing the advantage of GAN-based model.

摘要

•We add the adversarial model to image contour detection.•We employ all convolutional layers in the encoder stage to improve convergence rates.•We randomly crop 224 × 224 regions and then rescale them to improve fitting ability.•We evaluated the weight of adversarial loss showing the advantage of GAN-based model.

论文关键词：Contour detection,Deep learning,Generative adversarial networks,Convolutional neural networks

论文评审过程：Received 26 April 2018, Revised 17 August 2018, Accepted 20 September 2018, Available online 28 October 2018, Version of Record 19 December 2018.

论文官网地址：https://doi.org/10.1016/j.knosys.2018.09.033