The application provides a way to display images in a slideshow; all the images with incorrect orientation are auto adjusted. It uses Principal Component Analysis (PCA), Haar cascade classifier and takes into consideration factors like illumination, exposure to processes the image.
The MNIST database is an excellent source of handwritten digits. It contains 60,000 training images and 10,000 testing images. Typically, a convolutional net is used to get good results. I propose a quick online recognition, XOR based heuristic that works fairly well given the system’s quick response time.