This project is about building a convolutional neural network from scratch (using basic libraries like numpy, scipy and not using high performance TensorFlow libraries). After that, we used the network to develop an image classifier that can classify the handwritten digits from 0 to 9. Moreover, we can extend this classifier to classify any images after training the network.
As we said earlier, we developed the network from scratch. We didn't primarily use TensorFlow. Rather we used numpy, scipy etc and developed the network from ground. We did an extensive search of the hyperparameters to get the best possible hyperparameters.