Improving Accuracy in Object Detection Using Region-Based Convolutional Neural Network
T.M Kiran Kumar, Srinivas Aluvala, Preetam Mandal, K. Saranya, Karunakara Rai B · 2023
The object detection is a fundamental computer vision task that involves identifying and locating objects within a video or image. This technique employed various shape patterns for suggestion to recognize the certain objects in an input data. The existing methods have drawbacks like cannot identify the small region of objects in image. In this research, the Region-based Convolutional Neural Network (RCNN) is proposed for object detection. The dataset utilized for object detection is Caltech101 dataset and it is pre-processed by normalization and LAB color transformation to enhance the image quality. The preprocessed images are given into ResNet-50 based feature extraction and RCNN is utilized for classification. The performance of RCNN model is estimated through the performance measures of accuracy, precision, recall, specificity and f1-score. The RCNN attained high accuracy of 95.78%, precision of 94.81%, recall of 94.27%, specificity of 93.65% and f1-score of 92.49% which is comparatively superior than other existing methods like Histogram of Oriented Gradient (HOG) and Local Ternary Pattern (LTP), Hierarchical Max-poling (HMAX) and Fuzzy C-means.