An Automatic Object Detection and Location System applying Faster R-CNN

摘要

A Great effort has been done to solve the problem of object detection and recognition in computer vision. This paper describes the results of experiments on detection and recognition of objects in images by using the Faster Region Convolutional Neural Networks (Faster R-CNN) technique. The experiments were performed using a calibrated vision system, a set of 150 images containing a total of 4 different objects. The obtained accuracy for the detection and recognition task was satisfactory, varying from 90.13% to 100%. With these results, we can use the proposed method to estimate the joint configuration of a manipulator and allow it to grasp a desired object.