Abstract: In recent years, great progress has been made in the design of the flexible continuum arms, but its modeling research lags behind relatively. This paper focuses on the kinematic modeling of the flexible continuum arm, involving the inverse kinematics, which is hard to converge. The problem is solved by using a temporary pose matrix. The feedforward neural network is used to fit the model based on the mode shape function (MSF), and an end-to-end forward kinematics model and an inverse kinematics model are established, which not only ensures the accuracy of the model, but also significantly reduces the time in solving the inverse kinematics. In addition, the model allows a flexible change of the number of the input and output neurons of the feedforward neural network, thus helps to adapt to a multi-section model.