Skip to content
Go to file

Traditional Chinese Landscape Painting Dataset

Paper Title: "End-to-End Chinese Landscape Painting Creation Using Generative Adversarial Networks"
Current GAN-based art generation methods produce unoriginal artwork due to their dependence on conditional input. Here, we propose Sketch-And-Paint GAN (SAPGAN), the first model which generates Chinese landscape paintings from end to end, without conditional input. SAPGAN is composed of two GANs: SketchGAN for generation of edge maps, and PaintGAN for subsequent edge-to-painting translation. Our model is trained on a new dataset of traditional Chinese landscape paintings never before used for generative research. A 242-person Visual Turing Test study reveals that SAPGAN paintings are mistaken as human artwork with 55% frequency, significantly outperforming paintings from baseline GANs. Our work lays a groundwork for truly machine-original art generation.

Sketch-And-Paint GAN, compared with baseline models: Alt Text

Here, we provide the dataset used to train our Sketch-And-Paint GAN model. The dataset consists of 2,192 high-quality traditional Chinese landscape paintings (中国山水画). All paintings are sized 512x512, from the following sources:

For more details about dataset collection methodology, please see the paper.

Dataset Samples:
Alt Text

Please cite the paper if you choose to use this dataset for your research.

      title={End-to-End Chinese Landscape Painting Creation Using Generative Adversarial Networks}, 
      author={Alice Xue},


Dataset used for WACV 2021 paper: "End-to-End Chinese Landscape Painting Creation Using Generative Adversarial Networks"



No releases published


No packages published
You can’t perform that action at this time.