ialtamirano/ocr_demo

Optical Character Recognition - Demo implementation

Streamlit-based web application to perform OCR on input images

License

MIT

Installation

Install instructions

  git clone https://github.com/lamb-does-code/ocr_demo
  cd ocr_demo
  pip install -r requirements.txt

Run Locally

This project has been tested using Ubuntu 18.04.5

Open Streamlit webapp

streamlit run app.py

Launch very basic Tkinter UI

python gui.py 

The two mains

• main.py

  Parameters:

  • gpu: bool (default False)
  • img_path: required

python main.py --gpu --img_path

• main_video.py

  Parameters:

  • gpu: bool (default False)
  • video_path: required

python video_main.py --gpu --video_path

Usage/Examples

On the streamlit app:

Enable/Disable GPU usage accordingly. Choose an input image, the output will be the same image with highlighted boxes of your text, and a paragraph of the detected text with the relative boxes

On the two mains:

In main.py and main_video.py, the output will be the same as the streamlit app, but it will be saved in output/

You will have a folder of your experiment, where you'll find the text-highlighted image, plus two text files for your text and your boxes

On the main_video.py, the output will be a new video, with the high-lighted text, and a text file for the boxes.

On the tkinter UI:

As of now, it works with images only, and the output is the same as in main.py

Demo image