Image Processing & Computer Vision

Below is a listing of the work presented at the SIGGRAPH 2020 Posters. To explore the content in detail, please review the program’s listing on the ACM Digital Library.

View Content Details

Back to posters

Poster: Posters: Image Processing & Computer Vision
Event TypePoster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
6D Pose Estimation With Two-stream Net
SessionPosters: Image Processing & Computer Vision
Contributors
Xiaolong Yang
Xiaohong Jia
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionIn this Poster, we present a heterogeneous architecture with two-stream network for estimating 6D object pose from RGB images, and our results outperform state-of-the-art approaches in two public datasets.
Contributors
Xiaolong Yang
AMSS
UCAS
Xiaohong Jia
AMSS
UCAS
A Companding Based Two-layer Codec for HDR Images
SessionPosters: Image Processing & Computer Vision
Contributors
Asif Siddiq
Jameel Ahmed
Ishtiaq Rasool Khan
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionThere is something new in the old. Utilizing a veteran technique of companding, we design a two-layer codec for HDR images which outperforms the existing codecs in terms of accuracy and compression of data.
Contributors
Asif Siddiq
Riphah International University Islamabad
Jameel Ahmed
Riphah International University Islamabad
Ishtiaq Rasool Khan
University of Jeddah Jeddah
A Fast and Practical CNN Method for Artful Image Regeneration
SessionPosters: Image Processing & Computer Vision
Contributors
Xiaolin Wu
Qifan Gao
Zhenhao Li
Shenglei Li
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe observe that photo retouching can be modeled by a parametric, monotonically non-decreasing tone mapping function in luminance and by an affine transform in chrominance. This property allows us to reduce the complexity of existing CNN methods for photo retouching by two orders of magnitude and also improve their robustness.
Contributors
Xiaolin Wu
McMaster University
Qifan Gao
Shanghai Jiao Tong University
Zhenhao Li
Shanghai Jiao Tong University
Shenglei Li
Shanghai Jiao Tong University
Bound-constrained Optimized Dynamic Range Compression
SessionPosters: Image Processing & Computer Vision
Contributors
Dorian Y. Chan
James F. O'Brien
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe propose a new bound-constrained framework for high dynamic range compression that computes a globally optimal manipulation of input pixel differences that is free of spatial artifacts while preserving local detail.

FIRST PLACE - GRADUATE DIVISION - STUDENT RESEARCH COMPETITION
Contributors
Dorian Y. Chan
Carnegie Mellon University
James F. O'Brien
University of California, Berkeley
Computational Image Marking on Metals via Laser Induced Heating
SessionPosters: Image Processing & Computer Vision
Contributors
Sebastian Cucerca
Piotr Didyk
Hans-Peter Seidel
Vahid Babaei
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionThis work discovers the gamut of color laser marking using a multi-objective optimization and exploits it to mark full-color images on some metals.
Contributors
Sebastian Cucerca
Max-Planck-Institut für Informatik
Piotr Didyk
Università della Svizzera Italiana
Hans-Peter Seidel
Max-Planck-Institut für Informatik
Vahid Babaei
Max-Planck-Institut für Informatik
Correspondence Neural Network for Line Art Colorization
SessionPosters: Image Processing & Computer Vision
Contributors
Trung Dang
Thien Do
Anh Nguyen
Giao Nguyen
Van Pham
Quoc Nguyen
Bach Hoang
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe propose a deep architecture to colorize line arts based on a reference image by finding corresponding components. We extract, for each component, an embedding vector. These embeddings are used to determine component correspondences between the reference and a line art, so reference colors can be propagated onto corresponding regions.
Contributors
Trung Dang
Cinnamon AI
Thien Do
Cinnamon AI
Anh Nguyen
Cinnamon AI
Giao Nguyen
Cinnamon AI
Van Pham
Cinnamon AI
Quoc Nguyen
Cinnamon AI
Bach Hoang
Cinnamon AI
DeepFaceDrawing: Deep Generation of Face Images From Sketches
SessionPosters: Image Processing & Computer Vision
Contributors
Shu-Yu Chen
Wanchao Su
Lin Gao
Shihong Xia
Hongbo Fu
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe present an interactive system to produce high-quality face images from rough or incomplete freehand sketches by implicitly modeling the shape space of plausible face images.
Contributors
Shu-Yu Chen
Institute of Computing Technology, Chinese Academy of Sciences
Wanchao Su
City University of Hong Kong
Lin Gao
Institute of Computing Technology, Chinese Academy of Sciences
Shihong Xia
Institute of Computing Technology, Chinese Academy of Sciences
Hongbo Fu
City University of Hong Kong
Fast and Deep Facial Deformations
SessionPosters: Image Processing & Computer Vision
Contributors
Stephen W. Bailey
Dalton Omens
Paul DiLorenzo
James F. O'Brien
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe present a deep learning-based approximation method for complex film-quality facial rigs. Our method significantly decreases deformation evaluation time, and it provides a differentiable model for use with inverse kinematics.
Contributors
Stephen W. Bailey
University of California, Berkeley
Unity Technologies
Dalton Omens
University of California, Berkeley
Paul DiLorenzo
DreamWorks Animation
James F. O'Brien
University of California, Berkeley
GAN-based AI Drawing Board for Image Generation and Colorization
SessionPosters: Image Processing & Computer Vision
Contributors
Minghao Li
Yuchuan Gou
Bo Gong
Jing Xiao
Mei Han
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionBy incorporating colorization into generation via a novel and lightweight feature embedding method, we achieved a GAN-based drawing board which takes semantic and color tone inputs from users and automatically generates paintings, enabling creation of pictures or paintings with semantics and color control in real time.
Contributors
Minghao Li
PAII INC.
Yuchuan Gou
PAII INC.
Bo Gong
PAII INC.
Jing Xiao
Ping An Technology
Mei Han
PAII INC.
ICF: A Shape-based 3D Segmentation Method
SessionPosters: Image Processing & Computer Vision
Contributors
Koji Kobayashi
Takafumi Aoki
Koichi Ito
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe propose a new 3D segmentation method, Incremental Contour Flow, which divides a 3D object into segments separated by boundary surfaces at narrow parts. It utilizes the property of distance transform that generates local maxima in the center of the swollen part and bottleneck-like structures in the narrow part.
Contributors
Koji Kobayashi
Vocsis Corporation
Takafumi Aoki
Tohuku University
Koichi Ito
Tohoku University
Interferometric Transmission Probing with Coded Mutual Intensity
SessionPosters: Image Processing & Computer Vision
Contributors
Alankar Kotwal
Anat Levin
Ioannis Gkioulekas
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe introduce a programmable, interferometric technique to selectively probe light paths in a scene, with applications to tissue imaging and optical coherence tomography.
Contributors
Alankar Kotwal
Carnegie Mellon University
Anat Levin
Technion
Ioannis Gkioulekas
Carnegie Mellon University
Motion-attentive Network for Detecting Abnormal Situations in Surveillance Video
SessionPosters: Image Processing & Computer Vision
Contributors
U-Ju Gim
Jeong-Hun Kim
Kwan-Hee Yoo
Aziz Nasridinov
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe propose a motion-attentive network called MA-Net that learns enhanced deep-learned features and preserves significant portion of motion information related to abnormal situations in video frames. These enhanced deep-learned features solve the problem of failing to detect anomalies due to the loss of sparse information.
Contributors
U-Ju Gim
Chungbuk National University
Jeong-Hun Kim
Chungbuk National University
Kwan-Hee Yoo
Chungbuk National University
Aziz Nasridinov
Chungbuk National University
PoPGAN: Points to Plant Translation With Generative Adversarial Network
SessionPosters: Image Processing & Computer Vision
Contributors
Yuki Yamashita
Kenta Akita
Yuki Morimoto
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe propose a two-stage deep learning model which generates high-resolution plant images and animations using relatively little training data, and point-input method, to facilitate simple input.
Contributors
Yuki Yamashita
Kyushu University
Kenta Akita
Kyushu University
Yuki Morimoto
Kyushu University
Reflection Removal via Realistic Training Data Generation
SessionPosters: Image Processing & Computer Vision
Contributors
Youxin Pang
Mengke Yuan
Qiang Fu
Dong-Ming Yan
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe propose a valid polarization-based reflection contaminated image synthesis method, which provides an adequate, diverse, and authentic training dataset, enhances the neural network by introducing the reflection information as guidance, and utilizes adaptive convolution kernel size to fuse multiscale information, which significantly improves the single image reflection removal results.
Contributors
Youxin Pang
National Laboratory of Pattern Recognition (NLPR), Chinese Academy of Sciences
School of Artificial Intelligence, University of Chinese Academy of Sciences
Mengke Yuan
National Laboratory of Pattern Recognition (NLPR), Chinese Academy of Sciences
School of Artificial Intelligence, University of Chinese Academy of Sciences
Qiang Fu
King Abdullah University of Science and Technology (KAUST)
Dong-Ming Yan
National Laboratory of Pattern Recognition (NLPR), Chinese Academy of Sciences
School of Artificial Intelligence, University of Chinese Academy of Sciences
The Phenomenon of Eclipsed Bokeh
SessionPosters: Image Processing & Computer Vision
Contributor
Paul Debevec
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionOut-of-focus points of light, when obscured by out-of-focus occluders, become “eclipsed” by a sharply-focused traveling occlusion edge which can move at a speed different from that of the occluder, and even in the opposite direction. The phenomenon can produce interesting visual effects in photos and motion pictures.
Contributor
Paul Debevec
USC Institute for Creative Technologies (ICT)
Google Research
The Role of Objective and Subjective Measures in Material Similarity Learning
SessionPosters: Image Processing & Computer Vision
Contributors
Johanna Delanoy
Manuel Lagunas
Ignacio Galve
Diego Gutierrez
Ana Serrano
Roland Fleming
Belen Masia
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe compare several networks trained for material similarity and find that the internal structure of deep neural networks can model perceptual material similarity, even without training with human data.
Contributors
Johanna Delanoy
Universidad de Zaragoza
Manuel Lagunas
Universidad de Zaragoza
Ignacio Galve
Universidad de Zaragoza
Diego Gutierrez
Universidad de Zaragoza
Ana Serrano
Universidad de Zaragoza
Roland Fleming
Justus-Liebig-Universität Giessen
Belen Masia
Universidad de Zaragoza
Unsupervised Learning of Visual Representations by Solving Shuffled Long Video-frames Temporal Order Prediction
SessionPosters: Image Processing & Computer Vision
Contributors
Fatemeh Siar
Amin Gheibi
Ali Mohades
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe propose a model for learning visual representation by solving order prediction tasks. We concatenate the frame pairs instead of the feature pairs. This concatenation makes it possible to apply 3D-CNN to extract features from the frame pairs. We propose a new grouping, which has achieved 80% accuracy on average.
Contributors
Fatemeh Siar
Amirkabir University of Technology
Amin Gheibi
Amirkabir University of Technology
Ali Mohades
Amirkabir University of Technology
Using Convex Hull for Fast and Accurate Ellipse Detection
SessionPosters: Image Processing & Computer Vision
Contributors
Zeyu Shen
Mingyang Zhao
Xiaohong Jia
Dong-Ming Yan
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe present a novel method for fast and accurate ellipse detection based on an efficient arc grouping strategy. Our approach achieves promising results on both synthetic and three real-world datasets.
Contributors
Zeyu Shen
National Laboratory of Pattern Recognition (NLPR), Chinese Academy of Sciences
School of Artificial Intelligence, University of Chinese Academy of Sciences
Mingyang Zhao
KLMM, Academy of Mathematics and Systems Science, Chinese Academy of Sciences
University of Chinese Academy of Sciences
Xiaohong Jia
KLMM, Academy of Mathematics and Systems Science, Chinese Academy of Sciences
University of Chinese Academy of Sciences
Dong-Ming Yan
National Laboratory of Pattern Recognition (NLPR), Chinese Academy of Sciences
School of Artificial Intelligence, University of Chinese Academy of Sciences
Vid2Curve: Simultaneous Camera Motion Estimation and Thin Structure Reconstruction from an RGB Video
SessionPosters: Image Processing & Computer Vision
Contributors
Peng Wang
Lingjie Liu
Nenglun Chen
Hung-Kuo Chu
Christian Theobalt
Wenping Wang
Event Type
Poster
Interest Areas
Arts & Design
Research & Education
Keywords
AI/Machine Learning
Rendering
Time
LocationCarousel
DescriptionWe present a new framework to reconstruct complex thin objects with high quality.
Contributors
Peng Wang
University of Hong Kong
Lingjie Liu
Max-Planck-Institut für Informatik
Nenglun Chen
University of Hong Kong
Hung-Kuo Chu
National Tsing Hua University
Christian Theobalt
Max-Planck-Institut für Informatik
Wenping Wang
University of Hong Kong