Wednesday January 31, 2018
Keynote: Deep Learning for Recognition and Detection I
Session Chair: Qian Lin, HP Labs, HP Inc. (United States)
9:10 – 10:10 AM
Harbour A-B
IMAWM-310
How does building a low cost vision sensor teach us about deep learning?, Tianli Yu, Morpx Inc. (United States)
Dr. Tianli Yu is the CEO and co-founder of Morpx Inc., a startup based in Hangzhou that delivers innovative computer vision hardware and software. He received his PhD in ECE from the University of Illinois at Urbana Champaign (2006). After graduation, he's been a senior computer vision researcher in Motorola Labs working on the embedded stereo depth camera for Motorola's phones. Later, Dr. Yu joined like.com and designed algorithms to assist shoppers in finding their personal styles. Like.com was eventually acquired by Google in 2010. After working for a few years in the design of large scale visual search and recognition algorithms for Google Shopping, Dr. Yu founded Morpx with his friend Frank Ran in late 2013. Morpx is the second time in his career that he is working to build an ultra-compact and super energy efficient computer vision system.
Deep Learning has been proven to be very successful for many recognition problems in computer vision, but the reason for this success is still relatively unclear, largely due to its enormous computation and non-linearity hiding behind the seemingly simple stacked neural layers. These complexities also create practical problems for embedded applications where power and computation resources are extremely limited. Morpx Inc has been spending years on delivering low-cost / low-complexity visual recognition system. We believe that there are still large computational redundancies in the commonly used deep learning networks. In our lab, by scaling down the deep networks to bare minimum and test how different optimization techniques affect these networks’ performance, we gained some useful insights that cannot be observed on regular deep nets. In this talk, I will present some of these insights and a survey of many others from the global embedded vision community. We hope that these insights will shed some lights on the directions where a convincing theory for deep learning could later be developed.
10:00 AM – 4:00 PM Industry Exhibit
10:10 – 10:50 AM Coffee Break
Deep Learning for Recognition and Detection II
Session Chairs:
Andreas Savakis, Rochester Institute of Technology (United States) and Jane You, The Hong Kong Polytechnic University (Hong Kong)
10:50 AM – 12:30 PM
Harbour A-B
10:50IMAWM-336
Depth and super-pixel extraction for augmenting human detection (Invited), Theus Aspiras, Hussin Ragb, and Vijayan Asari, University of Dayton (United States)
11:30IMAWM-467
Vision based vehicle re-identification by fusion of multiple features, Geng Yang1, Jane You2, Zhenhua Guo3, and Qin Li1; 1Shenzhen Genvict Technologies Co., Ltd. (China), 2The Hong Kong Polytechnic University (Hong Kong), and 3Tsinghua University Shenzhen Graduate School (China)
11:50IMAWM-338
Hierarchical Auto-associative Polynomial Convolutional Neural Network (HAP-CNN) for pattern classification, Patrick Martell, University of Dayton Research Institute (United States)
12:10IMAWM-339
Learn a hybrid collaborative representation for fine-grained image classification, Wenyang Xie1, Bao-Di Liu1, Xue Li2, and Yan-Jiang Wang1; 1University of China University of Petroleum (Huadong) and 2Tsinghua University (China)
12:30 – 2:00 PM Lunch
2:00 – 3:00 PM PLENARY: Ubiquitous, Consumer AR Systems to Supplant Smartphones
3:00 – 3:30 PM Coffee Break
Deep Learning for Face Recognition
Session Chair:
Reiner Fageth, CEWE Stiftung & Co. KGAA (Germany)
3:30 – 5:10 PM
Harbour A-B
3:30IMAWM-372
One-shot face recognition: A review (Invited), Yandong Guo and Lei Zhang, Microsoft Research (United States)
4:10IMAWM-373
Face liveness detection based on joint analysis of RGB and near-infrared image of faces, Lingxue Song and Changsong Liu, Tsinghua University (China)
4:30IMAWM-374
CNN based facial landmark detection, Ruiyi Mao1, Qian Lin2, and Jan Allebach1; 1Purdue University and 2HP Labs, HP Inc. (United States)
4:50IMAWM-421
Empirical study of image compression for palm vein recognition, Zhenhua Guo1, Qin Li2, Yujiu Yang1, and Jane You3; 1Tsinghua University (China), 2Shenzhen Institute & Information Techology (China), and 3The Hong Kong Polytechnic University (Hong Kong)
5:30 – 7:00 PM EI 2018 Symposium Interactive Papers (Poster) Session
5:30 – 7:00 PM Meet the Future: A Showcase of Student and Young Professionals Research
Thursday February 1, 2018
Deep Learning for Recognition and Detection III
Session Chair:
Zhigang Fan, Apple Inc. (United States)
9:10 – 10:10 AM
Harbour A-B
9:10IMAWM-419
3D Shape Retrieval using volumetric and image convolutional neural networks: A meta-algorithmic approach, Ruiting Shao1, Yang Lei2, Jian Fan2, and Jerry Liu2; 1Purdue University and 2HP Inc. (United States)
9:30IMAWM-420
A new fast template matching algorithm for object detection, Jichao Jiao1, Weihua Tang2, Xin Wang1, and Zhongliang Deng1; 1Beijing University of Posts and Telecommunications and 2China State Construction Engineering Corporation (China)
9:50IMAWM-375
A feature fusion strategy for human detection in omnidirectional camera imagery, Hussin Ragb and Vijayan Asari, University of Dayton (United States)
10:10 – 10:50 AM Coffee Break
Multimedia Analytics in Online Systems
Session Chair:
Vijayan Asari, University of Dayton (United States)
10:50 AM – 12:30 PM
Harbour A-B
10:50IMAWM-443
Has mobile photography changed the users’ behavior while ordering printed products? (Invited), Reiner Fageth, CEWE Stiftung & Co. KGAA (Germany)
11:30IMAWM-444
Application of natural language processing to an online fashion marketplace, Kendal Norman1, Zhi Li1, Young-Taek Oh1, Gautam Golwala2, Sathya Sundaram2, and Jan Allebach1; 1Purdue University and 2Poshmark Inc. (United States)
11:50IMAWM-445
Multimedia analytics platform for profiling keywords embedded in photo catalogues, Emiliano Pallotti, Federica Mangiatordi, Andrea Bernardini, and Licia Capodiferro, Fondazione Ugo Bordoni (Italy)
12:10IMAWM-446
Use of color information in the analysis of fashion photographs, Zhi Li1, Gautam Golwala2, Sathya Sundaram2, and Jan Allebach1; 1Purdue University and 2Poshmark Inc. (United States)
12:30 – 2:00 PM Lunch
Mobile Image Analytics and Augmented Reality
Session Chair:
Yandong Guo, Microsoft Research (United States)
2:00 – 3:20 PM
Harbour A-B
2:00IMAWM-453
Semantic pose machines (Invited), Ying-Kai Huang and Andreas Savakis, Rochester Institute of Technology (United States)
2:40IMAWM-454
Learning enhancement with mobile augmented reality, Xunyu Pan and Joseph Shipway, Frostburg State University (United States)
3:00IMAWM-455
Person segmentation using convolutional neural networks with dilated convolutions, David Ho1 and Qian Lin2; 1Purdue University and 2HP Labs, HP Inc. (United States)
3:20 – 3:50 PM Coffee Break
Multi-Media Object Detection
Session Chair:
Jan Allebach, Purdue University (United States)
3:50 – 4:50 PM
Harbour A-B
3:50IMAWM-466
Deep learning for moving object detection from a single camera in UAVs (Invited), Dong Hye Ye, Qiulin Chen, Jing Li, Juan Wachs, and Charles Bouman, Purdue University (United States)
4:30IMAWM-337
Logo detection and recognition with synthetic images, Daniel Mas Montserrat1, Qian Lin2, Jan Allebach1, and Edward Delp1; 1Purdue University and 2HP Labs, HP Inc. (United States)