In his book, Prince argues that computer vision should be understood in terms of measurements (images), the world state, a model (defining the statistical relationships between the observations and the world), parameters, and learning and inference algorithms. The presented environment can address all these elements in modern computer vision R&D. Also, such a environment keeps evolving. For example, in the raising field of deep learning vision, packages as Keras, PyTorch, and Theano are promising tools that could become an important part of the Python ecosystem for scientific computing in a near future.
Questions?