Hybrid Convolutional Neural Network Model to ascertain the Objects in Dynamic Cluttered Environment

Abstract

The field of computer vision has made significant strides in object detection in recent years, primarily because of the introduction of deep learning techniques, specifically Convolutional Neural Networks (CNNs). We have introduced a novel method for the multi-object detection in multi-scene cluttered environment in the proposed work. In order to build multi-scale andmulti-scene object detection, in our work we have provides a multi-scale neural network basedonthehigherresponse of FastR-CNNarchitecture. For the experimental work,we have considered different categories of different objects. The dataset is designed to facilitate the development of object detection techniques. It comprises 12,165 object chips, each consisting of 256 pixels in both azimuth and range dimensions. This dataset encompasses diverse primary backgrounds and object sizes. Furthermore, well-known cutting-edge object detectors that have been trained on real-world images are modified to serve as baselines, guaranteeing the availability of reliable and practical reference points. Experimental results indicate that these object detectors not only enhance various quantitative metrics but also achieve unprecedented levels of accuracy, surpassing the capabilities observed in prior studies.