|
Kumar Chellapilla |
|
Welcome! I am a Researcher in the Document Processing and Understanding group in Microsoft Research. Short Bio: Kumar H. Chellapilla holds a Ph.D. in Electrical and Computer Engineering from University of California at San Diego. Before joining Microsoft, he worked a Senior Staff Scientist at Natural Selection, Inc. in La Jolla, CA. He is currently doing research on advanced technologies for document processing and information extraction. His research interests include machine learning, pattern recognition, computer vision, and computational intelligence. Some of his current projects are described below. For further references try the publications link. |
|
Research |
|
Human Interaction Proofs (papers: NIPS’04, CHI’05, HIP’05, CEAS’05) joint work with Patrice Simard Human interaction proofs (HIPs) are commonplace on the internet for protecting free online services from abuse by automated scripts/bots. They are challenges designed to be easily solved by humans, while remaining too hard for computers to solve. |
|
Document Processing and Understanding |
|
Combining Classifiers in a Cascade (papers: DRR’06, DAS’06) joint work with Patrice Simard & Michael Shilman Conventional approaches to combining classifiers improve accuracy at the cost of increased processing. We propose a novel search based approach to automatically combine multiple classifiers in a cascade to obtain the desired tradeoff between classification speed and classification accuracy. The search procedure only updates the rejection thresholds (one for each constituent classier) in the cascade, consequently no new classifiers are added and no training is necessary. Heuristic methods for finding approximate solutions and branch-and-bound search for optimal solutions are being explored. On handwritten/print character recognition such cascade classifiers can generate 5-10x speedup in processing times. |
|
Font Hashing (paper: ICDAR’05) joint work with Patrice Simard & Radoslav Nickolov A glyph hashing approach for fast optical character recognition is designed for use in document conversion applications. The hashing approach is used to not only recognize the character codes but also, size, style (bold, italic, etc), and font name. It can scale to hundreds of fonts and thousands of characters per font. Owing to its speed, such a hashing approach can complement any existing OCR system by acting as a pre-filter to produce a 4-5 times speedup during document conversion. |
|
BitHash: C5100C4A782... |
|
ContourHash: E1336C94B1F... |
|
ß |
|
ß |