CNE-Reader is a notebook that utilizes the TensorFlow 2 Object Detection API to train an SSD-MobileNet model and convert it to TensorFlow Lite format. Its purpose is to extract information from student ID cards issued by the National Institute of Posts and Telecommunications - INPT.
CNE-Reader comprises two main steps. Firstly, we employ Transfer Learning to train a model capable of localizing Matricule numbers and Full Names on the student ID cards. Subsequently, we utilize the PyTesseract library to extract text from the localized zones. For further details on the model, refer to the ocr_model notebook.
If you only intend to use the model, download requirements.txt and upload it to Colab. Then, execute !pip install -r requirements.txt. Alternatively, if you plan to run the model's source code, there is no need to install the requirements.
3. 1. Download the custom_model_lite folder.
3. 2. Open the model_test.ipynb notebook on Google Colab.
3. 3. Follow the steps outlined in the notebook.
The input images go through several preprocessing steps such as resizing, normalization and color correction before being fed into the model.
Utilizing OpenCV to process the localized images before feeding them to PyTesseract to read the text.
TensorFlow Object Detection API