02141nam a22001337a 4500 65493 65490 Siyal, Mehwish 15MSSE17 Supervisor - Dr.Fareed Ahmed Jokhio Image Category Classification Using Resnet50 a Pretrained Network Nawabshah: QUEST, 2019. 44p. ABSTRACT lmage classification currently is used to lesson the gap between the human vision and Computer Vision (CV) so that the image can be predicted by machines in the same mode as we human do. It accords with n signing the suitable class for the given certain image. Since the last decade, Deep Learning (DL) has paved the direction for demanding and wide-ranging applications in almost everyday life and accomplished good performance on several problems such as in the application area of natural language processing, face recognition, weather forecasting and health informatics. motion detection. Extending the structure of Neural Network s (NNs) Convolutional Neural Networks (CNNs) have been most expansively/commonly used architecture in Deep Learning. The research on convolutional neural network has been emerged speedily and attained state-of-the-art results. In this research work, a system is proposed that performs image category classification using Resnet50 ; pretrained CNNs model and furthermore, Resnet50 is used with SVM classifier and Transfer Learning. The extracted features from already trained CNNs are used to train a Support Vector Machines (SVM) classifier for classification and the fine-tuning of layers is achieved by retraining in the last few layers of Resnet50 under transfer learning for classification. The obtained results show a high accuracy of 95% on a classification task by taking up to 50 different categories from Caltech_ 101 dataset consists on maximum of two thousand five hundred images using pretrained network with SVM and Transfer Learning. Furthermore, while comparing, SYM classifier showed better performance due to reduced training time than Transfer Learning. Department of Information Technology http://tinyurl.com/56czuxxv THESIS 0 0 0 0 RESEARCH RESEARCH 2023-12-20 0 MP/54-686 2023-12-20 00:00:00 2023-12-20 THESIS 0 0 0 0 RESEARCH RESEARCH 2023-12-20 0 MP/46-545 2023-12-20 00:00:00 2023-12-20 THESIS