Using neural networks to recognize digits in a scene

Neural nets are widely used (as example for the MNIST dataset). Using neural networks and convolutional neural networks, in particular, we can get over 99% accuracy. However, the MNIST dataset is a dataset where each image only has one digit cropped and centered so that it is always in the same position.

If I were to create a neural network to recognize digits where the digits may not be in the center (and possibly there may be other objects or digits in the scene), how do I go about developing it?

I need to detect the digits and their position.