Understanding Images

  • Full Conference Pass (FC) Full Conference Pass (FC)
  • Full Conference One-Day Pass (1D) Full Conference One-Day Pass (1D)

Date: Thursday, December 6th
Time: 9:00am - 10:45am
Venue: G402 (4F, Glass Building)
Session Chair(s): Alex Kim, Magic Leap, United States of America


Gourmet Photography Dataset for Aesthetic Assessment of Food Images

Abstract: We present Gourmet Photograph Dataset, the first large-scale dataset for food photo aesthetics. We verify its effectiveness via extensive experiments with state-of-the-art visual machine learning algorithms and unseen food photos.

Authors/Presenter(s): Kekai Sheng, NLPR, Institute of Automation, Chinese Academy of Sciences; University of Chinese Academy of Sciences, China
Weiming Dong, NLPR, Institute of Automation, Chinese Academy of Sciences, China
Haibin Huang, Megvii/Face++ Research, United States of America
Chongyang Ma, Snap Inc., United States of America
Bao-Gang Hu, NLPR, Institute of Automation, Chinese Academy of Sciences, China


On the Convergence and Mode Collapse of GAN

Abstract: We present a novel architecture of GAN. The new architecture with a novel loss function improves the convergence problem and basically solves the mode collapse problem of the GAN.

Authors/Presenter(s): Zhaoyu Zhang, University of Science and Technology of China, China
Mengyan Li, University of Science and Technology of China, China
Jun Yu, University of Science and Technology of China, China


Removing Objects from Videos with A Few Strokes

Abstract: We propose a complete system for segmenting and removing chosen objects in videos, taking as only input hand-drawn approximate outlines of these objects in at least one frame.

Authors/Presenter(s): Thuc Trinh Le, LTCI, Telecom ParisTech; Paris-Saclay University, France
Andrés Almansa, MAP5, CNRS & Université Paris Descartes, France
Yann Gousseau, LTCI, Telecom ParisTech; Paris-Saclay University, France
Simon Masnou, Claude Bernard Lyon 1 University; Institut Camille Jordan, CNRS UMR 5208, France


Dunhuang Mural Restoration using Deep Learning

Abstract: We propose a systematic restoration process for high-resolution deteriorated mural textures, and show the potential for learning different image domain transfer with GAN.

Authors/Presenter(s): Han-Lei Wang, National Taiwan University, Taiwan
Ping-Hsuan Han, National Taiwan University, Taiwan
Yu-Mu Chen, National Taiwan University, Taiwan
Kuan-Wen Chen, National Chiao Tung University, Taiwan
XINYI LIN, National Taiwan University, Taiwan
Ming-Sui Lee, National Taiwan University, Taiwan
Yi-Ping Hung, National Taiwan University, Tainan National University of the Arts, Taiwan


 

Back

/en/attendees/virtual-augmented-reality-vr-ar /en/attendees/technical-briefs