Post-irradiation testing of nuclear fuel must be carried out in a special installation called a hot laboratory (Hotcell). The main component is the manipulator, which functions as a robotic arm that holds nuclear fuel samples in all of its displacement processes. As technology advances, manipulators can be operated automatically. The integration of a camera in the manipulator allows the application of computer vision technology in identifying the geometric shape of nuclear fuel, so that in the future a manipulator can be created that can run part of the testing process automatically. In this study, a system that can recognize the shape of a nuclear fuel plate in the form of PEB (Pelat Elemen Bakar) U3Si2Al has been developed. The algorithm approach was built using YOLO (You Only Look Once). The results obtained are that the geometric shape of the sample can be recognized with an accuracy value of 95% and can distinguish if the sample held by the manipulator is not nuclear fuel.