Pedestrian crossing prediction is a crucial task for autonomous driving.\nNumerous studies show that an early estimation of the pedestrian's intention\ncan decrease or even avoid a high percentage of accidents. In this paper,\ndifferent variations of a deep learning system are proposed to attempt to solve\nthis problem. The proposed models are composed of two parts: a CNN-based\nfeature extractor and an RNN module. All the models were trained and tested on\nthe JAAD dataset. The results obtained indicate that the choice of the features\nextraction method, the inclusion of additional variables such as pedestrian\ngaze direction and discrete orientation, and the chosen RNN type have a\nsignificant impact on the final performance.\n