/Curated-scene-text-recognition-analysis

Curated lists of scene text recognition and analysis

Curated-scene-text-recognition-analysis

Contributing

Please feel free to pull requests, email seungwoo.yoo@gmail.com.

Datasets

  1. Oxford Dataset Synthetic word / character datasets. Containing 9M images covering 90k English words. Introduced in the "M. Jadderberg, and Vedaldi, A. and Zisserman, A. Deep Features for Text Spotting, ECCV14"
  2. Street View Text Street view text image dataset. Gathering from Google Street View. Introduced in the "Kai Wang and Serge Belong Word Spotting in the Wild, ECCV 10"
  3. Chars74K Cropped character images including English, Latin and Hindu-Arabic numerals. Consisting of 64 classes, 7k characters from natural images, 3k handwritten, 60k synthetic images.
  4. ICDAR03 Robust Reading For ICDAR 2003 Robust Reading competitions.
  5. ICDAR05 Robust Reading For ICDAR 2005 Robust Reading competitions.
  6. MSRA Text Detection Containing 500 natural images, with various geometric distortion like perspective distortion. The resolutions of the images vary from 1296x864 to 1920x1280. Introduced in the "C. Yao, X. Bai, W. Liu, Y. Ma and Z. Tu. Detecting Texts of Arbitrary Orientations in Natural Images. CVPR 2012"
  7. SIW-10 Script identification datasets - 13,045 multi-scripts text line images in 10 classes, cropped from 7,700 full images taken in-the-wild (Google street view). Introduced in the "Baoguang Shi, Cong Yao, Chengquan Zhang, Xiaowei Guo, Feiyue Huang, Xiang Bai, Automatic Script Identification in the Wild. ICDAR 2015"
  8. Text/Nontext scene dataset Contains 5302 text images and 6000 non-text images, while the testing set contains 2000 text images and 2000 non-text images. Introduced in "Chengquan Zhang, Cong Yao, Baoguang Shi, Xiang Bai, Automatic Discrimination of Text and Non-Text Natural Images, ICDAR 2015"

Open-source

  1. [Tesseract] (https://github.com/tesseract-ocr/tesseract) Optical character recognition toolkit. Recognize 39 languages and UTF8 capable. Written in C++.
  2. clstm A small C++ implementation of LSTM networks, focused on OCR. Written in C++ with Python bindings.
  3. ocrpy Python-based OCR package using recurrent neural networks.
  4. ocrDemo OCR-implementation in Android using OpenCV and Tesseract
  5. LTPTextDetector Text detector written in C++. Introduced in "Michael Opitz, Markus Diem, Markus Diem, Florian Kleber, Stefan Fiel and Robert Sablatnig, End-to-End Text Recognition using Local Ternary Patterns, MSER and Deep Convolutional Neural Networks, DAS 2014.
  6. Vuforia Text Recognition Qualcomm Vuforia Text recognition SDK. Support custom word lists, various fonts and styles.

Miscellaneous

  1. IAPR TC11 Dataset List International Association for Pattern Recognition, Technical Committee 11. Well documented and summarized various datasets and tools for document recognition and analysis

Paper