Bearings, as core and vulnerable components in electric motor systems, play a crucial role in determining the motor's performance and lifespan. The real-time monitoring of bearing health is, therefore, a significant challenge. Current methods primarily rely on feature selection based on the internal principles of the motor bearings, utilizing the fault characteristic frequencies of the motor to make judgments. However, these methods are often inaccurate and heavily dependent on human expertise. In this paper, we propose an end-to-end real-time fault diagnosis algorithm for motor bearings, named MotorNN, which utilizes a deep neural network to perform fault diagnosis based on data, thereby avoiding the complexities of internal parameters of motor bearings and achieving high diagnostic accuracy. MotorNN consists of two modules: the Perception Module and the Decision Module. Temporal-domain signals are first processed through the Perception Module to extract relevant features, which are then classified using the Decision Module. Furthermore, we compare MotorNN with three other methods, demonstrating the superior performance of MotorNN.