This group project work is a Recurrent Neural Network with the joint implementation of two different Self-Supervised Tasks. The purpose of this RNN is to tackle the video recognition task, in terms of first person egocentric action recognition. The project is described with great details in the report.