This repository contains an Optical Character Recognition (OCR) system that combines two important components: text detection using the EAST (Efficient and Accurate Scene Text) model and character recognition using a custom model trained on the EMNIST dataset. The OCR system aims to accurately detect and recognize text from images.
- Character recognition using a custom model trained on EMNIST
- Keras , Tensorflow , open-cv,python
- Achieves a validation accuracy of 85.06 percent
- Open-cv EAST model or open -cv contour based model can be used for character segmenation. An example of both is provided
- Used Keras tuner to find the best hypermaramters.