Deep CNN Based Hybrid Model for Image Retrieval
Amit Sharma1, V.K. Singh2, Pushpendra Singh3
1Amit Sharma, Research Scholar, Motherhood University, Roorkee (Uttarakhand), India.
2Dr. V.K. Singh, Professor, Motherhood University, Roorkee (Uttarakhand), India.
3Dr. Pushpendra Singh, Raj Kumar Goel Institute of Technology, Ghaziabad (Uttar Pradesh), India.
Manuscript received on 09 July 2022 | Revised Manuscript received on 18 July 2022 | Manuscript Accepted on 15 August 2022 | Manuscript published on 30 August 2022 | PP: 23-28 | Volume-11 Issue-9, August 2022 | Retrieval Number: 100.1/ijitee.G92030811922 | DOI: 10.35940/ijitee.G9203.0811922
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The popularity of deep features based image retrieval and classification task has grown a lot in the recent years. Feature representation based on Convolutional Neural Networks (CNNs) found to be very effective in terms of accuracy by various researchers in the field of visual content based image retrieval. The features which are neutral to their domain knowledge with automatic learning capability from their images are in demand in various image applications. For improving accuracy and expressive power, pre-trained CNN models with the use of transfer learning can be utilized by training them on huge volume of datasets. In this paper, a hybrid model for image retrieval is being proposed by using pre-trained values of hyper parameters as input learning parameters. The performance of the model is being compared with existing pre-trained models showing higher performance on precision and recall parameters.
Keywords: Content based Image Retrieval, Deep Convolutional Neural Network, Transfer Learning, Pre-trained Models.
Scope of the Article: Convolutional Neural Network