Department of Computer Science
Rochester Institute of Technology
Phone: (585) 475-4536
|[ Home ]||[ Members ]||[ Projects ]||[ Publications ]||[ Software ]||[ Support ]|
The Document and Pattern Recognition Lab (dprl) is located in the Department of Computer Science at the Rochester Institute of Technology in Rochester, New York, USA. Our lab explores ways to improve the algorithms and tools used for automatic recognition and retrieval of information in documents, images, audio and video.
Math Information Retrieval (MIR) researchers at
SIGIR 2016 in Pisa, L-R: Richard Zanibbi, Kenny Davila,
Moritz Schubotz, Iadh Ounis, Bela Gipp, and Lingcai Gao.
Kenny Davila giving a talk for his paper at SIGIR (slides)
Dr. Lei Hu, May 2016
Dr. Siyu Zhu, May 2016
Top: Nidhin Pattaniyil @ NTCIR-11 in Tokyo.
Bottom: Kenny Davila @ ICFHR 2014 in Crete.
News (2016) (Earlier News)
- (July) Lei Hu and Prof. Zanibbi have had two papers on recognizing handwritten math accepted for publication at ICFHR, being held in Shenzhen, China this October. These papers summarize Lei's doctoral dissertation research.
- (July) The CROHME 2016 handwritten math recognition competition has ended! Congratulations to MyScript and Wiris for obtaining the highest recognition rates. Details of the competition will appear in an upcoming paper at ICFHR.
- (June) The new Tangent 0.3.1 formula search engine has been released (see Software link, above). This new version was a joint RIT-Univ. Waterloo collaboration, created by K. Davila, A. Kane, F.Wm. Tompa, and Prof. Zanibbi.
- (May) The NTCIR-12 MathIR Task Overview paper is now available. The paper describes the math search competitions held as part of the upcoming NTCIR-12 conference being held in Tokyo.
- (May) Kenny Davila, Andrew Kane, Frank Tompa and Prof. Zanibbi's paper about the Tangent math search engine at the NTCIR-12 math retrieval competitions is now available. For the competition we improved support for wildcard matching, and explored new metrics for formula similarity.
- (May) Congratulations to Yi Huang, who has accepted a position at BOSE corporation in Boston.
- (May) Congratulations to Kenny Davila who has successfully defended his PhD dissertation proposal entitled "Appearance-Based Retrieval of Mathematical Notation in Documents and Lecture Videos". The proposal builds upon Kenny's earlier research, including his recently accepted SIGIR 2016 paper describing modifications to the Tangent formula search engine.
- (Apr.) Congratulations to Dr. Siyu Zhu, who has successfully defended his doctoral dissertation entitled "Text Detection for Natural Scenes and Technical Diagrams with Convolutional Features and Cascaded Classification." Part of Siyu's doctoral research was recently accepted for publication at CVPR 2016. Siyu's innovative, but relatively simple text detection system obtained the highest rates for the standard ICDAR 2015 Focused Scene Text benchmark at the time of publication.
- (Apr.) Congratulations to Dr. Lei Hu, who has sucessfully defended his doctoral dissertation entitled "Features and Algorithms for Visual Parsing of Handwritten Mathematical Expressions." His dissertation presents new visual features and parsing algorithms (including parser ensembles) for recognizing handwritten mathematical notation, which illustrate that competitive recognition results may be obtained using very minimal language models - existing state-of-the-art approaches use expression grammars, which are defined and refined manually.
- (Apr.) Congratulations to Kenny Davila, who has received a travel award from the ACM to attend SIGIR 2016 in Pisa, Italy.
- (Apr.) Source code for Siyu Zhu's text detector for natural scenes is available from the Software link above.
- (Apr.) Congratulations (again) to Kenny Davila, who has been selected to participate in the Doctoral Consortium at SIGIR 2016.
- (Mar.) Congratulations to Kenny Davila, who co-authored a paper on scalable formula search entitled Multi-Stage Math Formula Search: Using Appearance-Based Similarity Metrics at Scale, which was accepted for publication at ACM SIGIR 2016 being held in Pisa, Italy. Kenny's co-authors include Frank Tompa and Andrew Kane from Univ. Waterloo, and Prof. Zanibbi. SIGIR is the leading international conference on Information Retrieval.
- (Mar.) Congratulations to Siyu Zhu, whose paper "A Text Detection System for Natural Scenes with Convolutional Feature Learning and Cascaded Classification" was accepted for publication at CVPR, the leading international computer vision conference. Siyu obtained state-of-the-art text detection results for the ICDAR 2015 Robust Reading Competition (Focused Scene Text Localization task).
- (Jan.) We welcome Yi Huang to the lab. Yi will create a web-based evaluation framework for handwritten math recognition systems, and look to extend this to support other recognition tasks (e.g. from computer vision and/or acoustics).
- (Dec. 2015) Congratulations to Kedarnath Calangutkar, whose poster tied for second place in the fall Master's project poster session. The title of his poster is Classification of Handwritten Math Symbols using Random Forest and Hybrid Features. In his work, Kedar combined pen/touch stroke features with shape based features previously used by Lei Hu and Kenny Davila and made modifications to the features, leading to an improved recognition rate.
- (Nov. 2015) A paper on unsupervised spoken word localization in lecture videos by Zack Miller, Manish Kanadje, Prof. Zanibbi and other members of the AccessMath project has been accepted for publication in Pattern Recognition Letters. The paper title is Assisted keyword indexing for lecture videos using unsupervised keyword spotting, and demonstrates the high accuracy of our system for both videos recorded at RIT, along with some Linear Algebra lectures recorded at MIT (demonstration).