Poshen Lee


Email: sephon@uw.edu | sephonlee@gmail.com
Go to my Linkedin
Go to my previous website
Review my Resume

Biography

Poshen Lee enrolled in University of Washington to pursue his Ph.D. degree in Electrical Engineering in 2011. He received his second M.S. from the same institude in 2013. Now he is working on VizioMetrics project with Prof. Bill Howe in Computer Science. This project aims at a novel scientific searching for schematic infomation. The techniques utilized in this project involve machine learning, computer vision, data anaylsis and data visualization. The VizioMetrics WebApp will be launched soon

Before coming to US, Poshen Lee had received his M.S. in Optics and Photonics and his B.S. in Physiscs both from National Central University, Taiwan. After graduate school, he complished his military service, and then joint his adviser, Prof. Mount Learn Wu's team to commercialize their silicon and optical system design technology in their start-up company, Centera Photonics Inc., as a technical cofounder. He was in charge of pico-projector development and also working on marketing strategy with Centera's CEO.

Research Interest

Data Scince, Computer Vision, and Human Computer Interaction

Specialties

Computer Science: Machine Learning, Data Visualization, Computer Vision, HCI
Art and Design: 3D modeling, 3D rendering, Graphic Design, Interior Design
Hardware Design: Lens design, Light Engine Design

Skill

Computer Programming: Java, Python, JavaScript, PHP, SQL, HTML, CSS, Matlab
Experienced Library: OpenCV, Amazon AWS API, D3 (Visualization), Android SDK, Django

Education

University of Washington, Seattle, U.S.A., 2011 - Present
Ph.D., Electrical Engineering

University of Washington, Seattle, U.S.A., 2011 - 2013
M.S., Electrical Engineering

National Central University, Jhongli, Taiwan, 2006 - 2008
M.S., Optics and Photonics

National Central University, Jhongli, Taiwan, 2002 - 2006
B.S., Physics

Researh Project

VizioMetrics
(Fall 2013 - present, RA in Computer Science Advised by Prof. Bill Howe and Prof. Linda Shapiro)

Scientific results are communicated visually in the literature through diagrams, visualizations, and photographs. These information-dense objects have been largely ignored in bibliometrics and scientometrics studies when compared to citations and text. In this project, we use techniques from computer vision and machine learning to classify more than 8 million figures from PubMed into 5 figure types and study the resulting patterns of visual information as they relate to impact. We find that the distribution of figures and figure types in the literature has remained relatively constant over time, but can vary widely across field and topic. We find a significant correlation between scientific impact and the use of visual information, where higher impact papers tend to include more diagrams, and to a lesser extent more plots and photographs. To explore these results and other ways of extracting this visual information, we have built a visual browser to illustrate the concept and explore design alternatives for supporting viziometric analysis and organizing visual information. We use these results to articulate a new research agenda – viziometrics – to study the organization and presentation of visual information in the scientific literature.

We originally used patch-based machine vision techniques to classify figures by visualization type, achieving 91% accuracy on a test set with 5 categories – equations (394), photos (782), tables (436), visualizations (890), and diagrams (769). More recently, we have begun using deep learning to achieve higher quality results at the expense of training time. For the task of classifying millions of images that we extracted from source papers, we found approximate 35% of them contains multiple sub-figures. A dismantling algorithm we proposed in ICPRAM 2015 resolves this issue by parsing each composite figure into multiple sub-figures. The algorithm splits each composite figure into visual “tokens” recursively, classifies each token as either auxiliary (e.g., text fragments) or standalone figures, then merges the tokens recursively to reconstruct whole figures. The algorithm terminates when the reonstructed figure achieve a certain “completeness” score based on their types and positions. Using the results of the dismantler, we can more precisely classify the sub-figures.

Visit our WebApp

Current Achievement:
1. Figure/Image Classifiers using Deep Learning
2. Composite Figure(Multi-chart figure) Dismantling Algorithm
3. Figure-centric Search Engine
4. Crowdsourcing Labbeller
5. Visual Pattern Analysis

On-going work:
1. Recover Phylogenetic Tree and Metabolic Pathway
2. Find The Most Important Figure in A Literature
3. VizioMetrics Open Data Platform
4. VizioMetrics Stats UI

Press and Recognition:
June 2016: The Economist has written a nice print piece on our arXiv paper.
June 2016: Top 5 tools of the week voted on LabWorm, a discovery platform that exposes top research tools with the goal of promoting a more open, collaborative and cutting edge science.
June 2016: MIT Technology Review wrote a nice piece on our project: The First Visual Search Engine for Scientific Diagrams

Papers:
1. "Dismantling Composite Visualizations in the Scientific Literature"
2. "VizioMetrix: A Platform for Analyzing the Visual Information in Big Scholarly Data"
3. "Viziometrics: Analyzing Visual Information in the Scientific Literature"

Related Project of Jevin West
1. Eigenfactor (eigenfactor.org)

Towards Streaming Summaries of High-Throughput Scientific Video (Summer 2014)
With the availability of novel sensor system, new experiments to observe fast physical or chemical reaction become possible under the frame rate of 1M fps. The significantly growing scale of video data requires developing new streaming algorithm even for common tasks such as retrieval, interactive browsing or compressing video content. We aim at developing summarization techniques to identify the most important and pertinent scientific content on the fly and produce a condensed version of video as an output. It can also benefit conventional methods by reducing the time spent on manually extracting important content from a long observation video. We believe that it could improve research productivity by saving time on low-productivity work.
View the poster

Mobile Accessibility for Vision-impared Users (Fall 2012 - Spring 2013, RA in Computer Science Advised by Prof. Richard Ladner)
The camera on smartphones and tablets are frequently used to capture images for future recall or for further retrieval. For blind people the camera can be very useful in finding out about their surroundings by optical character recognition (OCR) or human interpreting services. Unfortunately blind people may not take the best pictures. In some cases, a strong reflective light may damage information on the target object. In this project, we propose a reflection removal method specialized for object recognition aimed at digital displays, home appliance user interfaces, and documents with glossy surface. Our method additionally segments particular items, for instance buttons, displays, and text in a frame, as well as removes reflection on such items. Given two images taken from different viewpoints, our method modifies one image by selecting pixels or segments with weaker reflection from the other image.
This project was demonstrated in CVPR 2013.

Hello 911, Data Visualization for Criminal Calls in Seattle (Winter 2012)
Project Website
Go exploring the criminal records provided by Seattle government by using our online visualization tool!!! It is fun and very interesting !
Visualization for Weekly Crimes
Visualization for Yearly Crimes with Day and Night Length

CarbonShopper, Augmented Reality Application on Display Goggle with Head Camera (Spring 2012)
Prototype Video
Prototype Testing Video

Moving and Static Vehicles Traking (Spring 2012)
Example Video 1
Example Video 2

Joint Placement and Scheduling of Wireless Sensors (Winter 2011)
Achieved 40% improvement of sensing quality by using greedy algorithm to optimize number, location, and activated time of traffic sensors on California highway system

Pico-projector Imbedded in Mobile Phone (2010 - 2011)
In Centera Photonics Inc., as a optical design team leader, we were working on developing efficient and compact projector that is able to be imbedded in mobile devices. This project indcluded development of ultra-compact light engine and of ultra-compact projection lens.
Scheme

Selected Papers:
1. “The modulation of LEDs driving current and duration ratio in application of Color Sequential Pico-Projector”
2. “The Improvement of Efficiency and Uniformity in Non-image LED Illumination for Field-Sequential-Color Pico-projector”

Inter-chip Optical Interconnect (2007 - 2008, 2010 - 2011)
In NCU we were working on silicon-based 45° v-groove technology produced by wet etching. Prof. Mount-learn Wu from NCU led the same team to work on commercializing this technology on active optical cable such as QSFC and optical HDMI with 10 Gbps bandwidth.

Optical Modulation within Periodic Nano/Micro Structure (2006 - 2008)
We found and studied design rules on periodic Nano or Micro structure utilized to improve external quantum efficieny of LEDs, in other word, brightness of LEDs. An theoretical tool was built for the studied and was demonstrated by fabrications and measurements.

Selected Papers:
1. “Azimuthally isotropic irradiance of GaN-based light-emitting diodes with GaN microlens arrays”
2. “III-nitride-based microarray light-emitting diodes with enhanced light extraction efficiency”

Work Experience

University of Washington, Seattle, WA, USA, Fall 2012 - Present
Research Assistant in Computer Science

Pacific Northwest National Laboratory, Richland, WA, USA, Summer 2014
Research Assistant in Computer Science

Centera Photonics Inc., Hsinchu, Taiwan, April 2010 - July 2011
Technical Cofounder & Engineer

National Central University, Jhongli, Taiwan, December 2006 - June 2008
Research Assistant in Optics and Photonics

Arts and Design

Portfolio | The Selection of Design 2006 - 2009

Architecture Design | Pine Wave Guesthouse, National Central University

Architecture Design | Ripple Chapel

Architecture Design | Recluse Home

Product Design | Drink The Solar Power

Selected Honors

Candidate of Best Student Paper Award, selected to be included in the series "Lecture Notes in Computer Science", ICPRAM 2015

First Place Student Paper Award, Optics & Photonics Taiwan 2007

Award for Excellent Performance, Tic100 Talentrepreneur Innovation Collaboration 2007

Customer-Children Scholarship 2006, 2007 and 2008, Cathay Financial Group

Nan Shan Life Scholarship for Children of Policyholders 2007, Nan Shan Life Insurance Co.

Publication

Journal Paper

Po-Shen Lee, Jevin D. West, Bill Howe, "Viziometrics: Analyzing Visual Information in the Scientific Literature" In Prep.
view the paper

Baehr-Jones, Tom; Ding, Ran; Liu, Yang; Ayazi, Ali; Pinguet, Thierry; Harris, Nicholas C; Streshinsky, Matt; Lee, Poshen; Zhang, Yi; Lim, Andy Eu-Jin; Liow, Tsung-Yang; Teo, Selin Hwee-Gee; Lo, Guo-Qiang; Hochberg, Michael, “Ultralow drive voltage silicon traveling-wave modulator” Opt. Express Vol. 20, No.11, 12014-12020, 2012
view this paper

Mount-Learn Wu, Yun-Chih Lee, Shih-Pu Yang, Po-Shen Lee and Jenq-Yang Chang, “Azimuthally isotropic irradiance of GaN-based light-emitting diodes with GaN microlens arrays,” Opt. Express Vol. 17, No.8, 6148-6155, 2009
view this paper

Mount-Learn Wu, Yun-Chih Lee, Po-Shen Lee, Cheng-Huang Kuo, and Jenq-Yang Chang, “III-nitride-based microarray light-emitting diodes with enhanced light extraction efficiency,” Jpn. J. Appl. Phys. Vol. 47, No. 8, 6757-6759, 2008
view this paper

J.-W. Shi1, P-.Y. Chen, C.-C. Chen, J.-K. Sheu, W.-C. Lai, Yun-Chih Lee, Po-Shen Lee, Shih-Pu Yang, and Mount-Learn Wu, “Linear cascade GaN-based green light-emitting diodes with invariant high-speed/power performance under high-temperature operation,” IEEE Photon. Technol. Lett. Vol. 20, No. 23, 1896-1898, 2008
view this paper

Cheng-Wei Chien, Yun-Chih Lee, Po-Shen Lee, Jenq-Yang Chang and Jyh-Chen Chen,” Analysis of two-dimensional photonic bandgap structure fabricated by an interferometric lithographic system” Appl. Opt. Vol. 46, No.16, 3196-3204, 2007
view this paper

Conference Paper

Po-Shen Lee, Jevin D. West, Bill Howe, "VizioMetrix: A Platform for Analyzing the Visual Information in Big Scholarly Data" Proceedings of the 25th International Conference on World Wide Web (2016).
view the paper

Po-Shen Lee, Bill Howe, "Dismantling Composite Visualizations in the Scientific Literature" 4th International Conference on Pattern Recognition Applications and Methods (2015).
view the paper

Po-Shen Lee, Richard E. Ladner, "Homography-based Reflection Removal Specialized for Object Recognition on Mobile Platform" Demo in CVPR 2013.
view the poster

Chi-Yu Wang, Yun-Chih Lee, Po-Shen Lee, Chia-Hao Chiu, Cheng-Huang Kuo, Mount-Learn Wu, “Study of the optical effects of nanostructure embedded GaN light emitting diodes formed by nanorod template overgrowth” Optics and Photonics Taiwan 2010

view the poster

Chia-Hao Chiu, Sheng-Da Jiang, Po-Shen Lee, Yun-Chih. Lee, Wen-Shing Sun, and Mount-Learn Wu, “The modulation of LEDs driving current and duration ratio in application of Color Sequential Pico-Projector” Optics and Photonics Taiwan 2010
view this paper

Sheng-Da Jiang, Chia-Hao Chiu, Hsin-Chieh Wu, Po-Shen Lee, Yun-Chih Lee, Wen-Shing Sun, and Mount-Learn Wu, “The Improvement of Efficiency and Uniformity in Non-image LED Illumination for Field-Sequential-Color Pico-projector,” Optics and Photonics Taiwan 2010
view this paper

Y. C. Chang, C. C. Chang, H. C. Lan, J.W. Jheng, I. C. Lu, H. L. Hsiao, Y. C. Lee, Po-Shen Lee, and M. L. Wu, “Narrow Bandstop Filters Using Suspended Membrane Type of Guided-Mode Resonance Structures Based on Silicon,” Optics and Photonics Taiwan 2010
view this paper

Mount-Learn Wu, Yun-Chih Lee, Po-Shen Lee, Hsu-Liang Hsiao, Chia-Chi Chang and Jenq-Yang Chang, “Azimuthally concentrated irradiance of GaN-based light-emitting diodes with Si3N4 microstructure arrays,” in: Proc. MOC’09, Japan, Oct.2009
view this paper

Y. C. Lee, C. H. Hsu, S. P. Yang, Po-Shen Lee, M. L. Wu, and J.Y. Chang,”Modulation of uniform light pattern with light extraction enhancement by GaN microlens arrays of LEDs,” in: Proc. CLEO/QELS’ 09, USA, May. 2009
view this paper

Po-Shen Lee, Yun-Chih Lee, Cheng-Huang Kuo, Jenq-Yang Chang, Mount-Learn Wu, “Theoretical and experimental demonstration of enhanced light extraction efficiency in III-nitride-based micro-array light-emitting diodes,” in Proc. ODF’08 Taiwan, June. 2008
view this paper

Jenq-Yang Chang, Yun-Chih Lee, Po-Shen Lee, Cheng-Huang Kuo and Mount-Learn Wu “Mechanism of enhanced light extraction efficiency in III-nitride-based micro-array light-emitting diodes” in: Proc. MOC’07, Japan, Oct.2007
view this paper

Po-Shen Lee, Yun-Chih Lee, Cheng-Huang Kuo, Jenq-Yang Chang and Mount-Learn Wu, “Mechanism of enhanced light extraction efficiency in III-nitride-based micro-array light-emitting diodes,” in: Proc. Optics and Photonics Taiwan 2007
view this paper

Chien-Chi Hsu, Yun-Chih Lee, Shih-Pu Yang, Po-Shen Lee, Mount-Learn Wu, and Jenq-Yang Chang, "Modulation of uniform light pattern by using microstructure on p-GaN of LEDs," in: Proc. Optics and Photonics Taiwan 2007
view this paper