Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DLP-GAN: learning to draw modern Chinese landscape photos with generative adversarial network

Xiangquan Gui, Binxuan Zhang, Li Li, Yi Yang

TL;DR

This paper proposes DLP-GAN (Draw Modern Chinese Landscape Photos with Generative Adversarial Network), an unsupervised cross-domain image translation framework with a novel asymmetric cycle mapping, and introduces a generator based on a dense-fusion module to match different translation directions.

Abstract

Chinese landscape painting has a unique and artistic style, and its drawing technique is highly abstract in both the use of color and the realistic representation of objects. Previous methods focus on transferring from modern photos to ancient ink paintings. However, little attention has been paid to translating landscape paintings into modern photos. To solve such problems, in this paper, we (1) propose DLP-GAN (Draw Modern Chinese Landscape Photos with Generative Adversarial Network), an unsupervised cross-domain image translation framework with a novel asymmetric cycle mapping, and (2) introduce a generator based on a dense-fusion module to match different translation directions. Moreover, a dual-consistency loss is proposed to balance the realism and abstraction of model painting. In this way, our model can draw landscape photos and sketches in the modern sense. Finally, based on our collection of modern landscape and sketch datasets, we compare the images generated by our model with other benchmarks. Extensive experiments including user studies show that our model outperforms state-of-the-art methods.

DLP-GAN: learning to draw modern Chinese landscape photos with generative adversarial network

TL;DR

This paper proposes DLP-GAN (Draw Modern Chinese Landscape Photos with Generative Adversarial Network), an unsupervised cross-domain image translation framework with a novel asymmetric cycle mapping, and introduces a generator based on a dense-fusion module to match different translation directions.

Abstract

Chinese landscape painting has a unique and artistic style, and its drawing technique is highly abstract in both the use of color and the realistic representation of objects. Previous methods focus on transferring from modern photos to ancient ink paintings. However, little attention has been paid to translating landscape paintings into modern photos. To solve such problems, in this paper, we (1) propose DLP-GAN (Draw Modern Chinese Landscape Photos with Generative Adversarial Network), an unsupervised cross-domain image translation framework with a novel asymmetric cycle mapping, and (2) introduce a generator based on a dense-fusion module to match different translation directions. Moreover, a dual-consistency loss is proposed to balance the realism and abstraction of model painting. In this way, our model can draw landscape photos and sketches in the modern sense. Finally, based on our collection of modern landscape and sketch datasets, we compare the images generated by our model with other benchmarks. Extensive experiments including user studies show that our model outperforms state-of-the-art methods.
Paper Structure (20 sections, 10 equations, 13 figures, 5 tables)

This paper contains 20 sections, 10 equations, 13 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. GAN for image-to-image translation
  4. Style transfer for ancient Chinese painting
  5. Method
  6. Overview
  7. Network architectures
  8. Loss functions
  9. Experiments
  10. Data preparation
  11. Training details
  12. Qualitative results and analysis
  13. Landscape painting translation task
  14. Other translation tasks
  15. Quantitative evaluation
  16. ...and 5 more sections

Figures (13)

  • Figure 1: Example results of DLP-GAN in landscape painting style transfer. Top: ancient landscape painting to modern photos. Bottom: modern photos to ancient landscape painting. Left: landscape painting. Right: sketches.
  • Figure 2: Some example pictures collected for our style transfer tasks. (a) Ancient landscape paintings for ancient $\rightarrow$ modern. (b) Modern landscape photos for modern $\rightarrow$ ancient. (c) Landscape sketches for ancient $\rightarrow$ line drawing.
  • Figure 3: The pipeline exhibits the architecture of the proposed DLP-GAN. The pipeline consists of two directions operating different generators. Top: translation direction from Domain (X) Chinese landscape painting $\rightarrow$ Domain (Y) Modern photos. Bottom: Opposite direction.
  • Figure 4: Architecture of the generator $F$, in which k represents the kernel size, n is the number of feature maps and s is the stride in each convolutional layer.
  • Figure 5: Architecture of the generator $G$.
  • ...and 8 more figures