Robotic Arm Control By Fine-Tuned Convolutional Neural Network Model

Obtaining semantic information is crucial in order to implement complex robotic applications successfully. Therefore, it commonly expected from the robotics systems to be equipped with advanced hardware and software. In this study, the simulation results of a robotic arm, which manipulates the recognized objects using deep neural networks considering the physical features, are given for 10 different categories. An accuracy rate of %97.28 is achieved as a result of the fine-tuning of the deep neural network called VGGNet16 by using the dataset which is composed of 1000 training images and 400 testing images in each category. In addition, successful displacement results are obtained for all objects.