Rendering - Experimental Ideas & Implementations

Permanent URI for this community

https://diglib.eg.org/handle/10.2312/12635

Browse

Browsing Rendering - Experimental Ideas & Implementations by Subject "Applied computing"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    GAN-based Defect Image Generation for Imbalanced Defect Classification of OLED panels
    (The Eurographics Association, 2022) Jeon, Yongmoon; Kim, Haneol; Lee, Hyeona; Jo, Seonghoon; Kim, Jaewon; Ghosh, Abhijeet; Wei, Li-Yi
    Image classification based on neural networks has been widely explored in machine learning and most research have focused on developing more efficient and accurate network models for given image dataset mostly over natural scene. However, industrial image data have different features with natural scene images in shape of target objects, background patterns, and color. Additionally, data imbalance is one of the most challenging problems to degrade classification accuracy for industrial images. This paper proposes a novel GAN-based image generation method to improve classification accuracy for defect images of OLED panels. We validate our method can synthetically generate defect images of OLED panels and classification accuracy can be improved by training minor classes with the generated defect images.
  • Thumbnail Image
    Item
    SkyGAN: Towards Realistic Cloud Imagery for Image Based Lighting
    (The Eurographics Association, 2022) Mirbauer, Martin; Rittig, Tobias; Iser, Tomáš; Krivánek, Jaroslav; Šikudová, Elena; Ghosh, Abhijeet; Wei, Li-Yi
    Achieving photorealism when rendering virtual scenes in movies or architecture visualizations often depends on providing a realistic illumination and background. Typically, spherical environment maps serve both as a natural light source from the Sun and the sky, and as a background with clouds and a horizon. In practice, the input is either a static high-resolution HDR photograph manually captured on location in real conditions, or an analytical clear sky model that is dynamic, but cannot model clouds. Our approach bridges these two limited paradigms: a user can control the sun position and cloud coverage ratio, and generate a realistically looking environment map for these conditions. It is a hybrid data-driven analytical model based on a modified state-of-the-art GAN architecture, which is trained on matching pairs of physically-accurate clear sky radiance and HDR fisheye photographs of clouds. We demonstrate our results on renders of outdoor scenes under varying time, date, and cloud covers.