ABSTRACT
Convolutional Neural networks (CNN) have addressed many computer vision problems like image classification, object detection, object recognition etc. Many object detection algorithms detect an object present in an image by drawing bounding box around that object to localize that particular object using different CNN architectures and have achieved good results. When images are used as background to display text,the arrangement of text in relation to image can make image not to be so visually appealing. So far object detection and recognition algorithms focused on finding out specific object either in an image or video,But our paper mainly focuses on finding out an empty area in a background image using Faster R-CNN architecture so that user can enter text inside empty area without affecting the saliency part of image and also making image to look visually appealing. Finally, we demonstrate the results for empty region detection with average accuracy of 98%.
