Skip to content

Latest commit

 

History

History
18 lines (13 loc) · 632 Bytes

README.md

File metadata and controls

18 lines (13 loc) · 632 Bytes

ATR: Attentional Text Recognizer

This project is based on ASTER improvements to suit actual use.

TODO list

  • Basic version of Attentional Text Recognizer
  • Variable Length Support
  • Chinese Support
  • More flexible training data input pipeline to reduce the error bring by variable length
  • Beam Search
  • Bidirectional Decoder
  • Rectification network

Environments && Installation

  1. Go to c_ops and run build.sh to build the string_split_utf8 custom operator
  2. Add /path/to/atr to PYTHONPATH, or set this variable for every run