Adaptive and Versatile Document Understanding

in
PhD Seminar Course on

Adaptive and Versatile Document Understanding

Cagliari, July, 15-17, 2008
Instructor: Prof. Henry Baird
CSE Dept., Lehigh University, Bethlehem, Pennsylvania
Co-director of the Lehigh Pattern Recognition Research Laboratory. Co-founder of the Document Analysis Systems, Document Image Analysis for Libraries, and the Human Interactive Proofs workshop series. Fellow, IEEE and IAPR
Duration: 8 hours
Schedule:
  • Tuesday 15, 15-18
  • Wednesday 16, 10-13
  • Thursday 17, 10-12
Venue: Aula Y
Topics:
  1. Document Image Understanding for Digital Libraries:

    Web-based digital libraries: rapid expansion worldwide. Automatic document image understanding state of the art

  2. Document Content Extraction:

    Highly versatile document processing. Pixel-accurate segmentation and iterated classification

  3. Combination of Classifers:

    There is no dominant classifier. Combinations can be better; stochastic discrimination

  4. Contextual Analysis as an Aid to Recognition:

    Geometric, linguistic, and semantic context. The promise of computable semantics: reading Chess

  5. Adaptive Recognition Systems I:

    Self-retraining classifiers. Autonomously adaptive classifiers

  6. Adaptive Recognition Systems II:

    Exploiting isogeny (internal consistency) of documents. Towards whole-book recognition: anytime classifiers

  7. Human Interactive Proofs I:

    Survey of CAPTCHAs and other HIPs. The arms race: segmentation, occlusion, and linguistics

  8. Human Interactive Proofs II:

    Research frontier: exploiting style consistency and linguistic familiarity. Future directions: implicit CAPTCHAs, non-textual challenges

slides
Organizer: Prof. Fabio Roli
Dep. of Electrical and Electronic Engineering
University of Cagliari, Italy
Email: roli(at)diee.unica.it