OCR for Army Officer Evaluation Records
- Tutorial: https://medium.com/analytics-vidhya/optical-character-recognition-ocr-using-py-tesseract-part-1-29ba8104eb2b
- Remember to install tesseract: https://guides.library.illinois.edu/c.php?g=347520&p=4121425
- Test regex: https://regex101.com/
├── README.md <- The top-level README for users of this project.
│
├── data
│ ├── images <- Put your OER images here.
│ ├── text <- Text within images is saved here.
│ └── output <- The final, structured json is saved here
│
└── src
├── extract_from_oer.sh <- Creates json from one OER image.
├── extract_from_2_pg_oer.sh <- Creates json from two OER images.
├── split_images.py <- Splits one image into two.
├── binarize_images.py <- Binarizes images to black and white.
├── img_to_text.py <- Converts image text to .txt file.
└── parse.py <- Structures key text fields into json.
- git clone this repo
- save any images of OERs into
data/images/ - cd into src
- If one image with two pages (e.g. 2 pg pdf converted to .tiff)
- Format must be name.extension
- Call
bash extract_from_oer.sh [filename.extension] [threshold]
- If two images in two files
- Format must be name.page1.extension and name.page2.extension
- Call
bash extract_from_2_pg_oer.sh [file1.extension] [file2.extension] [threshold]
- Threshold is required for binarization of images into black and white
- split_images.py
- Split multi-page image into separate images. (good if turned pdf > image)
- python split_images.py [img_filename]
- binarize_images.py
- For making images more crisp
- python binarize_images.py [theshold 0-255] [img_file1] [img_file2] [file3...]
- img_to_text.py
- Extract text from images with tesseract + python wrapper
- Only include images/pages from same document (page 1 and 2 of OER)
- python img_to_text.py [oer_img_page1] [oer_img_page2]
- parse.py
- Extract/clean raw text and ouput a json file
- python parse.py [txt_filename]