The programming of the welding robot is a challenging problem, especially with complex paths. Extracting path points and suitable welding speed at every path zone is a complicated, time wasting, and costly process. Moreover, the accuracy of extracting these data at the design stage is affected by the inaccuracies in prewelding processes. This paper introduces a new supervised learning technique for programming a 4 degree of freedom (DOF) welding arm robot with automatic feeding electrode. In this technique, a three-dimensional (3D) machine vision system is developed to grasp the welding position and speed of a complex path by monitoring of an expert welding instructor. Then, these data are used to generate the robot move program. The proposed technique includes fewer steps and hence less consumed time than the conventional one. Moreover, it does not need an expert programmer. From the accuracy point of view, there is no significant difference between the two techniques. These enhancements will improve the share of robots in welding and similar industries.