Vision system development for hybrid robot-gripper to perform manipulation tasks in the food industry for object detection and localization

In this paper, it introduced how to apply and set up the computer vision system for the food packaging production line, in which this computer vision system based on the deep learning algorithm to develop. The paper had introduced and compared three of the current famous deep learning frameworks, which are TensorFlow, PyTorch, and Darknet. At the same time, it introduced and compared two of the different deep learning algorithms, which are the You Only Look Once (YOLO) and the Single Shot MultiBox Detector (SSD). Besides, the paper had demonstrated how to reproduce the YOLO and SSD model training procedures based on the PyTorch framework. In addition, the report demonstrated and discussed their actual object detecting results.Bachelor of Engineering (Mechanical Engineering