Digit_Recognition_CNN - Model Optimization using Batch Normalization and Dropout Techniques

In this project, various neural network architectures are successfully explored for digit recognition. The results demonstrate the effectiveness of convolutional neural networks in improving accuracy and generalization. Batch normalization and regularization techniques further enhance the model's performance. A high-performing CNN model is developed in this project that recognizes the digits accurately. At first, the multi-layers neural network is developed where the predictions are not that accurate. Then, the CNN model is developed and trained where the performance is a little bit better than the first model, but it is not that good. Furthermore, the CNN model is implemented with batch normalization and dropout layers to optimize its performance where this model can predict the digits very accurately with an accuracy of almost 99%.