Utilized two datasets :
1.The standard MNIST 0-9 dataset by LeCun et al.
2. The Kaggle A-Z dataset by Sachin Patel, based on the NIST Special Database 19
Trained the OCR model using Keras, TensorFlow, and deep learning architecture, ResNet.
Used OpenCV for image pre-processing (improve the image quality by removing noise and enhancing the contrast between the handwritten text and the background).
Keras is a high-level neural networks API built on top of TensorFlow, used Keras to provide a range of pre-built layers that can are easily combined.It is written in Python and is used to make the implementation of neural networks easy
TensorFlow can handle large datasets and is used to scale models across multiple devices, making it easier to train large models.
The TensorFlow platform helps you implement best practices for data automation, model tracking, performance monitoring, and model retraining.
ResNet uses a particular block called a residual block that allows the model to learn features from images, making it possible to create much deeper networks without encountering problems with vanishing gradients.
For downloading a_z_handwritten_data.csv using link :-
https://www.kaggle.com/datasets/sachinpatel21/az-handwritten-alphabets-in-csv-format and paste the csv file in the root directory.