[robocup-worldwide] IJCAI-PRICAI 2020 3D AI Challenge

Fri Jun 12 02:38:28 EDT 2020

IJCAI-PRICAI 2020 3D AI Challenge & 3D-FUTURE: 3D FUrniture shape with
TextUREMotivation, Impact, and Expected Outcomes

The vision community has put tremendous efforts into 3D object modeling
over the past decade, and has achieved many impressive breakthroughs.
However, there is still a large gap between the current state of the art
and industrial needs for 3D vision. One of the major reasons is that
existing 3D shape benchmarks provide only low-quality 3D shapes with
dreamlike or no textures. Another imperfection is that there is no large-
scale dataset in which 3D models exactly match the objects in images. These
are insufficient for comprehensive and subtle research in areas such as
texture recovery and transfer, which is required to support industrial
productions.

The goal of the AI challenge is to conclude state-of-the-art 3D geometry
and vision algorithms, to facilitate innovative research on high-quality 3D
shape understanding and generation, and to build a bridge between academic
research and 3D applications in industry. To support this goal, Alibaba
group has released *3D-FUTURE (3D FUrniture shape with TextURE): *a
richly-annotated, large-scale repository of 3D furniture shapes specific to
the household scenario. 3D-FUTURE contains 20,000+ clean and realistic
synthetic images in 5,000+ diverse rooms which involve 10,000+ unique
detailed 3D instances of furniture with high-resolution informative
textures.

[image: image.png]
Figure 1 3D-FUTURE Dataset

Highlights

   1. 1. High-quality shapes, Informative textures, Rich attributes

   The 3D shapes offered by public 3D benchmarks may show two
   imperfections. Firstly, most of these 3D CAD models (for furniture) are
   both with fewer details and low informative textures since they are
   collected online. Secondly, there are no diverse professional attributes
   for their furniture shapes. In contrast, 3D-FUTURE provides high-quality 3D
   furniture with rich details in various styles, including European furniture
   that often contains intricate carvings. Furthermore, each 3D shape in 3D-
   FUTURE is assigned to an informative texture and different attribute
   labels. We believe these features can potentially facilitate innovative
   research on high-quality 3D shape understanding and generation.
   [image: image.png]

   Figure 2 High-quality shapes & informative textures
   2.

   1. 2. Realistic renderings, real 2D-3D Alignment

There are no well-organized benchmarks that provide realistic synthetic
indoor images. 3D-FUTURE fill the blank by rendering 20,000+ images across
5,000+ scenes via one of the most advanced industrial 3D renders (V-Ray).
These indoor scenes are reviewed by professional designers. Besides,
existing benchmarks only provide pseudo 2D-3D alignment annotations.
Namely, they manually choose a roughly matched 3D CAD model from public 3D
shape benchmarks according to the object contained in the image. Annotators
thus may largely ignore some local shape details. As a result, these
benchmarks offer less matched 3D shape and 2D image pairs. This is not
sufficient to support data-driven studies such as high-quality 3D
reconstruction and high-accuracy 3D shape retrieval. Luckily, the 10,000+
3D shapes in 3D-FUTURE exactly match objects contained in the rendered
images.
[image: image.png]

Figure 3 realistic renderings & 2D-3D alignment
Competition tracks

Building on 3D-FUTURE, Alibaba 3D Artificial Intelligence Challenge 2020
has three tracks: 1) Cross-domain image-based 3D shape retrieval, 2) 3D
object reconstruction, and 3) Instance segmentation.

   1. 1. Cross-domain image-based 3D shape retrieval

   In this challenge, participants are required to retrieval the
   corresponding 3D shape given a 2D query image. We expect to foster the
   development of shape retrieval methods that are robust to slight occlusions
   and changes in diverse complicated surroundings. The performance will be
   mainly measured by TopK Recall.
   [image: image.png]

   Figure 4 image-based 3D shape retrieval
   2. 2. 3D object reconstruction

   In this challenge, participants will reconstruct a 3D shape from a
   single RGB image. The objects contained in the input RGB images may
   slightly occluded or partially

   incomplete. The Chamfer Distance and F-score will be used to measure the
   quality of the reconstruction.

   [image: image.png]

   3. 3. Instance segmentation

Figure 5 3D object reconstruction

In this challenge, participants are required to label each foreground pixel
with the appropriate object and instance. We include this challenge because
it will motivate vision-based image generation, thereby improving relevant
industrial production chains, for example by partially reducing the
requirement for expensive 3D rendering.

[image: image.png]

Infrastructures

Figure 6 Instance segmentation

Tianchi platform ( <http://tianchi.aliyun.com/)>http://tianchi.aliyun.com/):
We plan to use the data competition platform Tianchi developed by aliCloud
(part of the Group). Since 2014, 43 competitions have been successfully
hosted on Tianchi, which gathered 128,480 players from 77 countries and
regions. This platform is well developed, tested and can be tailored to
this contest as needed.
Competition Schedule

March 30 – July 24, 2020

Website URL

   1.

   The 3D AI challenge 2020 homepage:
   https://tianchi.aliyun.com/specials/promotion/ijcai-alibaba-3d-future-workshop
   2.

   The image-based 3D shape retrieval competition webpage:
   https://tianchi.aliyun.com/competition/entrance/231789/introduction
   3.

   The 3D object reconstruction webpage:
   https://tianchi.aliyun.com/competition/entrance/231788/introduction
   4.

   The Instance segmentation:
   https://tianchi.aliyun.com/competition/entrance/231787/introduction

OrganizersHuan Fu is currently a Senior Algorithm Engineer, Alibaba Group.
He received the PhD degree in computer science from The University of
Sydney in 2019. His research interests include scene understanding,
generative adversarial models, and 3D modeling from images. He has
published several research papers in CVPR and won the 1st prize in CVPR 3D
robust vision challenge.Rongfei Jia received the PhD degree in computer
science from Beihang University. He is currently a Staff Algorithm
Engineer, Alibaba Group. He leads an algorithm team which works on
innovative research in building home AI. His research interests include 3D
computer vision and data mining.Lin Gao received a PhD degree in computer
science from Tsinghua University. He is currently working as associate
professor and PhD supervisor at the Institute of Computing Technology,
Chinese Academy of Sciences. He won the Newton Advanced Fellowship award
from the Royal Society. His research interests include geometric
processing, computer vision and computer graphics.Mingming Gong is a
Lecturer (Assistant Professor) with the School of Mathematics and
Statistics, University of Melbourne. His research interests include
computer vision and machine learning. He has co-authored 30+ papers
published in ECCV, ICCV, CVPR, NeurIPS, ICML, etc. He has served as senior
program committee members of IJCAI and AAAI. He is an organizer of SDM
workshop on “Weakly-Supervised and Unsupervised Learning”.Binqiang Zhao is
a Senior Staff Engineer of the Taobao Technology Department of Alibaba
Group. He graduated from Tsinghua University in 2006. He is leading an
algorithm team which is devoted in 3D AI algorithms in furniture and
household industry and intelligent commercial platform operation. His team
has developed various algorithms for top user products in Tabao, such as
Tangping, Yangtao, Good Goods, and Taobao Live. His interests include 3D
shape modeling and 3D scene understanding.Steve Maybank is Professor of
Computer Science at Birkbeck College, University of London. His research
interests include camera calibration, visual surveillance, tracking,
filtering, applications of projective geometry to computer vision and
applications of probability, statistics and information theory to computer
vision. He is the author or co- author of more than 200 scientific
publications and one book. He is a Fellow of the IEEE and a Member of the
Academia Europaea. He received the Koenderink Prize in 2008. He was
co-chair and organiser for the following workshops: The IEEE International
Workshops on Visual Surveillance (1998, 1999, 2000, 2006, 2007, 2008); The
IEEE VS-PETS International Workshop on Visual Surveillance and Performance
Evaluation of Tracking and Surveillance (2003, 2005, 2006, 2007); The Rank
Prize Funds Mini-Symposium on Model Selection and Learning in Computer
Vision (2000); and BMVA Workshop 'New Approaches to Dynamic Filtering'
(1997).Dacheng Tao is Professor of Computer Science and ARC Laureate Fellow
in the School of Computer Science and the Faculty of Engineering, and the
Inaugural Director of the UBTECH Sydney Artificial Intelligence Centre, at
The University of Sydney. His research results in artificial intelligence
have expounded in one monograph and 200+ publications at prestigious
journals and prominent conferences, such as IEEE T- PAMI, IJCV, JMLR, AAAI,
IJCAI, NIPS, ICML, CVPR, ICCV, ECCV, ICDM, and

KDD, with several best paper awards. He received the 2018 IEEE ICDM
Research Contributions Award and the 2015 Australian Scopus-Eureka prize.
He is a Fellow of the IEEE, ACM and Australian Academy of Science. He was
an organiser for 10+ international conferences and workshops, such as ACM
CIKM 2019 (1000+ submissions, general co-chair) and IEEE ICDM 2018 (1000+
submissions, program co- chair).

Bibliographic references

   1.

   The 3D AI challenge 2020:
   https://tianchi.aliyun.com/specials/promotion/ijcai-
   alibaba-3d-future-workshop
   2.

   IJCAI-PRICAI 2020 3D AI Challenge – Baeslines, and Evaluation codes:
   https://github.com/FUTURE-3D/FUTURE3D-AI-Challenge-Baseline
   3.

   3D toolbox for texture rendering, 3D-2D projection, and 2D-3D
   reprojection: https://github.com/FUTURE-3D/FUTURE3D-ToolBox
   4.

   Tianchi Competition Platform: https://tianchi.aliyun.com/
   5.

   Alibaba group:  <http://www.alibabagroup.com/>
   http://www.alibabagroup.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-worldwide/attachments/20200612/7440e58e/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 196769 bytes
Desc: not available
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-worldwide/attachments/20200612/7440e58e/attachment-0006.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 147052 bytes
Desc: not available
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-worldwide/attachments/20200612/7440e58e/attachment-0007.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 162438 bytes
Desc: not available
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-worldwide/attachments/20200612/7440e58e/attachment-0008.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 119335 bytes
Desc: not available
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-worldwide/attachments/20200612/7440e58e/attachment-0009.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 98321 bytes
Desc: not available
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-worldwide/attachments/20200612/7440e58e/attachment-0010.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 118277 bytes
Desc: not available
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-worldwide/attachments/20200612/7440e58e/attachment-0011.png>