GAMES Webinar 2020 – 149期(CVPR专题) | Shangzhe Wu(University of Oxford), Yinyu Nie(Bournemouth University), Weipeng Xu(Facebook Reality Labs)
【GAMES Webinar 2020-149期】(CVPR专题)
报告嘉宾1：Shangzhe Wu(University of Oxford)
报告时间：2020年7月30号星期四晚上8:00 – 8:30（北京时间）
报告题目：Unsupervised Learning of 3D Objects from Images
In this talk, I will present our recent CVPR 2020 paper titled “Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild”. Training 3D reconstruction systems typically requires heavy supervision from various types of annotations, including 3D ground truth, multi-view images, depth maps, silhouettes, keypoints, camera viewpoints etc, which can be hard to obtain for all kinds of objects in our world. In this paper, we propose an unsupervised method for learning deformable 3D objects, which leverages a weak symmetry assumption. This method is able to learn accurate 3D shapes as well as intrinsic image decomposition simply from an unconstrained image collection without external supervision. This paper received the Best Paper Award at CVPR 2020.
Shangzhe Wu is a second year PhD student in Visual Geometry Group, University of Oxford, supervised by Prof. Andrea Vedaldi. His research has been focused on unsupervised 3D understanding. Prior to joining Oxford, Shangzhe received his bachelor’s degree from HKUST, where he worked with Prof. Chi-Keung Tang and Prof. Yu-Wing Tai on image translation.
报告嘉宾2： Yinyu Nie(Bournemouth University)
报告时间：2020年7月30号星期四晚上8:30 – 9:00（北京时间）
报告题目：Total3DUnderstanding: Joint Layout, Object Pose and Mesh Reconstruction for Indoor Scenes from a Single Image
Semantic reconstruction of indoor scenes refers to both scene understanding and object reconstruction. Existing works either address one part of this problem or focus on independent objects. In this paper, we bridge the gap between understanding and reconstruction, and propose an end-to-end solution to jointly reconstruct room layout, object bounding boxes and meshes from a single image. Instead of separately resolving scene understanding and object reconstruction, our method builds upon a holistic scene context and proposes a coarse-to-fine hierarchy with three components: 1. room layout with camera pose; 2. 3D object bounding boxes; 3. object meshes. We argue that understanding the context of each component can assist the task of parsing the others, which enables joint understanding and reconstruction. The experiments on the SUN RGBD and Pix3D datasets demonstrate that our method consistently outperforms existing methods in indoor layout estimation, 3D object detection and mesh reconstruction.
Yinyu Nie is currently a Ph.D. student at the National Centre for Computer Animation, Bournemouth University, UK. He received his MSc and BSc degree from Southwest Jiaotong University, China. His research interest focuses on 3D scene understanding, reconstruction and 3D shape analysis.
报告嘉宾3： Weipeng Xu(Facebook Reality Labs)
报告时间：2020年7月30号星期四晚上9:00 – 9:30（北京时间）
报告题目：DeepCap: Monocular Human Performance Capture Using Weak Supervision
Human performance capture is a highly important computer vision problem with many applications in movie production and virtual/augmented reality. Many previous performance capture approaches either required expensive multi-view setups or did not recover dense space-time coherent geometry with frame-to-frame correspondences. We propose a novel deep learning approach for monocular dense human performance capture. Our method is trained in a weakly supervised manner based on multi-view supervision completely removing the need for training data with 3D ground truth annotations. The network architecture is based on two separate networks that disentangle the task into a pose estimation and a non-rigid surface deformation step. Extensive qualitative and quantitative evaluations show that our approach outperforms the state of the art in terms of quality and robustness.
Weipeng Xu is a research scientist at Facebook Reality Labs in Pittsburgh. He was a post-doctoral researcher at the Graphic, Vision & Video group of Max Planck Institute for Informatics in Saarbrücken, Germany. He received B.E. and Ph.D. degrees from Beijing Institute of Technology in 2009 and 2016, respectively. He studied as a long-term visiting student at NICTA and Australian National University from 2013 to 2015. His research interests include virtual human character, human pose estimation and machine learning for vision/graphics.
GAMES主页的“使用教程”中有 “如何观看GAMES Webinar直播？”及“如何加入GAMES微信群？”的信息；