An Optimal Approach for Multi-class Object Detection

Abstract

Object detection in an image aims to point out all the desired objects in the target image and considers the positional enlightenment to obtain machine perspective knowledge. In object detection vanishing gradient has been the most significant problem that can occur, it results in a model with many layers being specified to unable to learn on a specific dataset. It could cause models to a low-grade solution. To get grip on the obstacles, this paper has analyzed certain algorithms for precise object detection from a customized dataset. In this approach, multiple objects in an image have been classified and localized with the help of Mask Region Colvolutional Neural Network (R-CNN). This work is pulled off using a regional convolutional neural network, pixellib, OpenCV, and TensorFlow, resulting in a preferable desired output. Implementing this technique and algorithm, based on deep learning, which is also based on machine learning requires framework understanding. The experimental result and analysis are done with multiple classes of an image to verify the efficiency of the mask R-CNN technique. The algorithm has been compared with other existing object detection strategies to establish the accuracy and reliability of the concerned determined algorithm. The systematic execution enabled clarification of the suggested technique to be precise in analyzing multi-class object in an image.