An Adaptive Convolutional Neural Network for Object Classification

Abstract

Since object detection has so many practical applications in areas as diverse as image retrieval, medical diagnosis, autonomous vehicle systems, surveillance, and more, it is one of the most investigated areas of automation. As the scope of use has grown, so too have the difficulties that must be overcome. This growth is due to the current models and architectures’ incapacity to identify obscured or diminutive items inside a picture. Images with low contrast or underexposure could benefit from sharpening or contrast restoration, or the input resolution could be increased to better detect small details. Alternately, the input resolution could be boosted to enhance large-object detection. Using multiple models for object detection has the potential to boost overall performance. The suggested study of the CDNN model aims to speed up object detection in challenging environments by decreasing the amount of time spent processing data. A custom-built deep convolutional neural network with additional augmentation layers has been proposed to take the place of the previous model’s backbone. It was demonstrated that the proposed model vastly increased the precision of