According to the International Federation of Robotics (IFR), the average robot usage density has grown to one robot per 10,000 employees in manufacturing. Robot usage is especially high in automotive, electronics, and metal industries. Rapid line change and 24-hour continuous operation have been major goals in robotic automation. However, current loading and unloading robots still rely heavily on structured environment and human tuning, thus resulting in long line change time. To remedy the issue, we propose an artificial intelligence-based algorithm to perform object pose estimation and robot picking or randomly stacked parts using 6 DOF robots. This algorithm first performs a learning-based object segmentation and then performs a 3D object pose estimation and grasping point determination.