RESUMEN
Content-Based Image Retrieval (CBIR) is the cornerstone of today's image retrieval systems. The most distinctive retrieval approach used, involves the submission of an image-based query whereby the system is used in the extraction of visual characteristics like the shape, color, and texture from the images. Examination of the characteristics is done for ensuring the searching and retrieval of proportional images from the image database. Majority of the datasets utilized for retrieval lean towards to comprise colored images. The colored images are regarded as in RGB (Red, Green, Blue) form. Most colored images use the RGB image for classifying the images. The research presents the transformation of RGB to other color spaces, extraction of features using different color spaces techniques, Gabor filter and use Convolutional Neural Networks for retrieval to find the most efficient combination. The model is also known as Gabor Convolution Network. Even though the notion of the Gabor filter being induced in CNN has been suggested earlier, this work introduces an entirely different and very simple Gabor-based CNN which produces high recognition efficiency. In this paper, Gabor Convolutional Networks (GCNs or GaborNet), with different color spaces are used to examine which combination is efficient to retrieve natural images. An extensive experiment using Cifar 10 dataset was made and comparison of simple CNN, ResNet 50 and GCN model was also made. The models were evaluated through a several statistical analysis based on accuracy, precision, recall, F-Score, area under the curve (AUC), and receiving operating characteristic (ROC) curve. The results shows GaborNet model effectively retrieve images with 99.68% of AUC and 99.09% of Recall. The results also shows different images are effectively retrieved using different color space. Therefore research concluded it is very significance to transform images to different color space and use GaborNet for effective retrieval.
RESUMEN
Computational visual encoding models play a key role in understanding the stimulus-response characteristics of neuronal populations in the brain visual cortex. However, building such models typically faces challenges in the effective construction of non-linear feature spaces to fit the neuronal responses. In this work, we propose the GaborNet visual encoding (GaborNet-VE) model, a novel end-to-end encoding model for the visual ventral stream. This model comprises a Gabor convolutional layer, two regular convolutional layers, and a fully connected layer. The key design principle for the GaborNet-VE model is to replace regular convolutional kernels in the first convolutional layer with Gabor kernels with learnable parameters. One GaborNet-VE model efficiently and simultaneously encodes all voxels in one region of interest of functional magnetic resonance imaging data. The experimental results show that the proposed model achieves state-of-the-art prediction performance for the primary visual cortex. Moreover, the visualizations demonstrate the regularity of the region of interest fitting to the visual features and the estimated receptive fields. These results suggest that the lightweight region-based GaborNet-VE model based on combining handcrafted and deep learning features exhibits good expressiveness and biological interpretability.