The ability to communicate is one of the most important rights of a healthy individual. Hearing-impaired citizens face several challenges: There is not a common sign language and the percentage of the population that understands a given sign language is negligible. The benefits of a system capable of translating from a sign language, in particular, the Mexican Sign language, to text, is outstanding. In this contribution, we describe such system. It is based on convolutional neural networks, which are particularly useful tool to extract relevant features from images. The sign language is captured in hundreds of images, and the system is able to translate them to text via several convolutional stages and transfer learning with 91% accuracy. We report the architecture and performance of the system.