Abstract
In the world of computer science, image recognition is a vital area that uses complex algorithms to make machines able to understand and classify visual data. The technology has applications in many fields including artificial intelligence, computer vision and others as it offers solutions for automated systems, security and object detection. This paper uses LaPlace GAN for image clarity enhancement and SSD (Single Shot MultiBox Detector) for effective object detection using one shot learning. By employing the strengths of LaPlace GAN, the model concentrates on improvement of intricate details and minimization of noise that leads to significant increase in overall clarity of an image. In addition, integration of SSD allows real time object detection hence the system can identify multiple objects simultaneously within a single step. Through training images from both PASCAL VOC2007 and PASCAL VOC2012 datasets, this research shows improved model performance. For training, proposed framework combines union of PASCAL VOC2007 train and VOC2012 trainval images; validation set is composed of PASCAL VOC2007val images while testing will be done with Pascal’s 2007 test images. The assessment gauge contains the Average Precision (AP) and mean Average Precision (mAP) calculated for all 20 classes that have been annotated in line with the PASCAL VOC object detection evaluation protocol.