We are proposing an OCR framework for recognizing ticker text in Russian Videos. We do this by solving two main problems, improving the OCR by implementing a Recurrent Neural Network, with a 1-dimensional (1-d) Bi-Directional Long Short-Term Memory (BDLSTM) architecture for Tensorflow which gives better results than Tesseract, and we solve the Scrolling Ticker problem by proposing an algorithm (Early Fusion) which checks the tickers and combine them in correct order.

Organization

Student

Burhan Ul Tayyab

Mentors

  • Anna Pleshakova
  • Abhinav Shukla
  • ivan.giangreco
  • Weixin Li
close

2018