9TH – 11TH DECEMBER 2019
SEMINAR HALL – IIT DELHI, INDIA
RECOGNITION AND RETRIEVAL OF DOCUMENTS IN INDIAN SCRIPTS
C. V. JAWAHAR
C. V. Jawahar is the Amazon Chair professor at IIIT Hyderabad, India. At IIIT Hyderabad, Jawahar leads a group focusing on AI, computer vision, and machine learning. In the recent years, he has been looking into the perception and learning problems that overlap with vision and language. He is also interested in designing AI systems, and applications with special focus on Indian setting. He is also passionate about the scalable models for teaching and research.
C.V. Jawahar has contributed significantly to visual recognition, especially in the area of text recognition, person recognition, fine grain classification, and structured prediction. He has worked extensively on recognizing text in Indian languages for OCRs and applications. Another area where Jawahar has been looking into is in enabling computer vision and machine learning for improving the Indian driving situations. A public data set (IDD) for this has been released in 2018. He is looking into the effectiveness of computervision for inspecting road conditions, improving safety and providing driver assistance. He is also interested in applications of AI inassistive technologies, healthcare, education, cultural heritage andentertainment.
He has more than 50 publications in top tier conferences in computer vision, robotics and document image processing. He has served as a chair for previous editions of CVPR, ICCV, AAAI, ACCV, WACV, IJCAI, ICDAR and ICVGIP. Presently, he is an area editor of CVIU andan associate editor of IEEE PAMI. He was a program co-chair for ACCV 2018, ICDAR 2017 and general chair for ICVGIP 2018 and ICFHR 2022. He is a Fellow of IAPR and INAE.
OCRs allow the content level access to the scanned or digitized document collections. However, these technologies are not yet fully available for Indian languages for deployment. In this talk, we discuss the associated challenges, state of the art and recent developments. Presentation will take examples of a recent attempt in recognizing a large collection of Indian language content (in three languages). In addition to the conventional OCRs, we also discuss a set of associated developments representation learning that can help in search or information access in large document image collections. Recent advances in machine learning and especially deep learning have opened up new promising directions in learning effective and efficient representations. Beyond the recognition of printed text in Indian languages, we also discuss the state and directions for the recognition in handwritten documents.
Recognition and Retrieval of Documents in Indian Scripts