Hardware Realization of Handwritten Text Detection

This study focuses on the implementation and evaluation of Hand Written Text Recognition (OCR) technology, employing the CRNN. The primary objective is to assess the performance of the OCR system, The model used is CRNN model in which CNN is used for feature extraction and LSTM is used for sequence prediction followed by a CTC layer which trained and has generated features The research involves a meticulous process of implementing and configuring the CRNN a hybrid model of CNN and Bidirectional LSTM. Subsequently, a comprehensive evaluation is conducted based on the accuracy of other models of the OCR system. Results obtained from the evaluation provide insights into the performance of the implemented OCR system, shedding light on its efficacy in accurately recognizing characters from various types of input images. The integration of CNN and LSTM has produced promising accuracy of 94.17%, paving the way for enhanced OCR capabilities on edge computing devices. This model is tested on NVIDIA® Jetson NanoTM Developer kit and the result is same as on PC. This research contributes valuable findings to the field of OCR technology, particularly in the context of edge computing platforms, and sets the stage for further advancements in character recognition systems.