Image-Based Visual Servoing (IBVS) is one of the popular approaches in visual servoing (VS) for robot manipulators by not requiring pose estimation. Besides this popularity, IBVS has to deal with two common problems in realization: obtaining the inverse of the interaction matrix and finding an appropriate fixed gain value for the controller. Although the interaction matrix for IBVS is used with pseudoinverse, the control law is not applicable in the case of singularities. On the other hand, fixed gain value causes a trade-off between convergence speed and endeffector velocities. In this study, an intelligent IBVS scheme is proposed to solve these problems. As the first stage of the system, the interaction matrix is replaced with a trained neural network and the singularity problem has been solved. Furthermore, the discontinuity of the initial velocities caused by the classical velocity controller are resolved by the used continuous velocity controller. As the second stage, instead of a fixed gain, a fuzzy logic unit inspired by fuzzy sliding mode and computing a gain value according to error and error derivative values in each loop is considered. Fast convergence without high velocity demand is provided by this adaptive gain approach.