2017 Western New York Image and Signal Processing Workshop
The 20th annual Western New York Image and Signal Processing Workshop (WNYISPW) is a venue for promoting image and signal processing research in our area and for facilitating interaction between academic researchers, industry researchers, and students. The workshop comprises both oral and poster presentations.
Date and Time
Location
Hosts
Registration
- Date: 17 Nov 2017
- Time: 08:30 AM to 05:30 PM
- All times are (GMT-05:00) US/Eastern
-
Add Event to Calendar
- Rochester Institute of Technology
- Rochester, New York
- United States
- Building: SLA/078
- Contact Event Host
- Co-sponsored by Signal Processing Society
- Starts 31 August 2017 02:00 AM
- Ends 16 November 2017 05:30 PM
- All times are (GMT-05:00) US/Eastern
- Admission fee ?
Speakers
Gao Huang
Efficient Training and Inference of Very Deep Convolutional Networks
Recent years have witnessed astonishing progress in convolutional neural networks (CNN). This fast development was largely due to the availability of training and inference of very deep models, which have even managed to surpass human-level performance on many vision tasks. However, the requirements for real world applications differ from those necessary to win competitions, as the computational efficiency becomes a major concern in practice. In the first part of the talk, I will introduce a stochastic depth network, which can significantly speed up the training process of deep models, making them more robust and generalize better. In the second part I will propose a densely connected network (DenseNet), which is inspired by the insights that we obtained from the stochastic depth network. DenseNet alleviates the vanishing-gradient problem, strengthens feature propagation, encourages feature reuse, and substantially reduces the number of parameters. The third part of the talk will focus on how to inference deep models under limited computational resources. I will introduce a multi-scale dense network (MSDNet) with shortcut classifiers, which facilitate retrieving fast and accurate predictions from intermediate layers, leading to significantly improved efficiency over state-of-the-art convolutional networks.
Biography:
Gao Huang is a postdoc researcher from the Department of Computer Science at Cornell University. He received the Ph.D. degree from the Department of Automation at Tsinghua University in 2015, and he was an intern/visiting scholar at Microsoft Research Asia (MSRA), Washington University in St. Louis and Nanyang Technological University. His research interests lie in machine learning and computer vision, with a focus on deep learning algorithms and network architectures. His paper “Densely Connected Convolutional Networks” won the Best Paper Award at CVPR 2017.
David J. Crandall
Egocentric Computer Vision, for Fun and for Science
The typical datasets we use to train and test computer vision algorithms consist of millions of The typical datasets we use to train and test computer vision algorithms consist of millions of consumer-style photos. But this imagery is significantly different from what humans actually see as they go about our daily lives. Low-cost, light wearable cameras (like GoPro) now make it possible to record people's lives from a first-person, "egocentric" perspective that approximates their actual fields of view. What new applications are possible with these devices? How can computer vision contribute to and benefit from this embodied perspective on the world? What could mining datasets of first-person imagery reveal about ourselves and about the world in general? In this talk, I'll describe recent work investigating these questions, focusing on two lines of work on egocentric imagery as examples. The first is for consumer applications, where our goal is to develop automated classifiers to help organize first-person images across several dimensions. The second is an interdisciplinary project using computer vision with wearable cameras to study parent-child interactions in order to better understand child learning. Despite the different goals, these applications share common themes of robustly recognizing image content in noisy, highly dynamic, unstructured imagery.
Biography:
David Crandall is an Associate Professor in the School of Informatics and Computing at Indiana David Crandall is an Associate Professor in the School of Informatics and Computing at Indiana University Bloomington, where he is a member of the programs in Computer Science, Informatics, Cognitive Science, and Data Science, and of the Center for Complex Networks and Systems Research. He received the Ph.D. in computer science from Cornell University in 2008 and the M.S. and B.S. degrees in computer science and engineering from the Pennsylvania State University in 2001. He was a Postdoctoral Research Associate at Cornell from 2008-2010, and a Senior Research Scientist with Eastman Kodak Company from 2001-2003. He has received an NSF CAREER award, a Google Faculty Research Award, best paper awards or nominations at CVPR, CHI, ICDL, ICCV, and WWW, and an Indiana University Trustees Teaching Award.