WICED 2022

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/2633159

Browse

Browsing WICED 2022 by Subject "Computing methodologies"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Consistent Multi- and Single-View HDR-Image Reconstruction from Single Exposures
    (The Eurographics Association, 2022) Mohan, Aditya; Zhang, Jing; Cozot, Remi; Loscos, Celine; Ronfard, Rémi; Wu, Hui-Yin
    Recently, there have been attempts to obtain high-dynamic range (HDR) images from single exposures and efforts to reconstruct multi-view HDR images using multiple input exposures. However, there have not been any attempts to reconstruct multi-view HDR images from multi-view Single Exposures to the best of our knowledge. We present a two-step methodology to obtain color consistent multi-view HDR reconstructions from single-exposure multi-view low-dynamic-range (LDR) Images. We define a new combination of the Mean Absolute Error and Multi-Scale Structural Similarity Index loss functions to train a network to reconstruct an HDR image from an LDR one. Once trained we use this network to multi-view input. When tested on single images, the outputs achieve competitive results with the state-of-the-art. Quantitative and qualitative metrics applied to our results and to the state-of-the-art show that our HDR expansion is better than others while maintaining similar qualitative reconstruction results. We also demonstrate that applying this network on multi-view images ensures coherence throughout the generated grid of HDR images.
  • Thumbnail Image
    Item
    Evaluation of Deep Pose Detectors for Automatic Analysis of Film Style
    (The Eurographics Association, 2022) Wu, Hui-Yin; Nguyen, Luan; Tabei, Yoldoz; Sassatelli, Lucile; Ronfard, Rémi; Wu, Hui-Yin
    Identifying human characters and how they are portrayed on-screen is inherently linked to how we perceive and interpret the story and artistic value of visual media. Building computational models sensible towards story will thus require a formal representation of the character. Yet this kind of data is complex and tedious to annotate on a large scale. Human pose estimation (HPE) can facilitate this task, to identify features such as position, size, and movement that can be transformed into input to machine learning models, and enable higher artistic and storytelling interpretation. However, current HPE methods operate mainly on non-professional image content, with no comprehensive evaluation of their performance on artistic film. Our goal in this paper is thus to evaluate the performance of HPE methods on artistic film content. We first propose a formal representation of the character based on cinematography theory, then sample and annotate 2700 images from three datasets with this representation, one of which we introduce to the community. An in-depth analysis is then conducted to measure the general performance of two recent HPE methods on metrics of precision and recall for character detection , and to examine the impact of cinematographic style. From these findings, we highlight the advantages of HPE for automated film analysis, and propose future directions to improve their performance on artistic film content.