An Efficient Image Classification using SVM, CNN

At first glance, the prospect of instructing a computer to perform tasks such as image classification seems highly captivating to our team.Furthermore, there exist several practical applications of this notion in various real-world scenarios.Based on these rationales, our research focus has been directed towards the field of Image Classification.Fortunately, this subject has been extensively investigated by the scientific community, and encountered little difficulty in locating resources for study.Consequently,thoroughly examined a multitude of scholarly articles pertaining to the subject of image classification, each offering a distinct viewpoint.Subsequently, the study made the determination to execute picture classification on a reduced scale due to the constraints imposed by the restricted technology at disposal.Despite the inherent challenges, the study initiated our analysis by employing Support Vector Machines (SVM) in conjunction with a rather limited dataset, ultimately attaining a commendable accuracy rate of 93%.While Support Vector Machines (SVM) is widely recognised as a robust technology, attaining exceptionally high accuracy remains an exceptional occurrence.It was observed that the great accuracy of our results can be attributed to the insufficiency of a sufficiently large dataset.By employing data augmentation techniques, the size of our dataset was expanded by more than threefold.When doing Support Vector Machine (SVM) analysis a notable reduction in accuracy is noted, with a recorded value of 81%.Dissatisfied with the outcomes, the recent works need to explore alternative deep learning methodologies.This inquiry directed our attention towards the study of Neural Networks, namely Convolutional Neural Networks (CNN).By effectively applying Convolutional Neural Networks (CNN), an impressive accuracy of 94.57% was attained on the identical dataset.This serves as evidence of the enhanced capabilities of deep learning approaches compared to conventional machine learning techniques.