Analysis of Content Based Image Retrieval using Deep Feature Extraction and Similarity Matching

Image retrieval using a textual query becomes major challenge mainly due to human perception subjectivity and the impreciseness of image annotations. These drawbacks can be overcome by focusing on content images rather than descriptions images. Traditional feature extraction techniques demand for expert knowledge select limited types are also sensitive changing imaging conditions. Deep Convolutional Neural Network (CNN) solution these as they learn representations automatically. This work carries out detailed performance comparison various pre-trained models CNN in extraction. Features extracted from men footwear women clothing datasets VGG16, VGG19, InceptionV3, Xception ResNet50 models. Further, features used classification SVM, Random Forest K-Nearest Neighbors classifiers. Results show that Inception perform well with extraction, achieving good accuracy 97.5%. results further justified performing efficiency, similarity metrics. compares obtained selected pretrained conventional CIFAR 10 dataset. The image-based systems like recommender systems, where have analyzed generate item profiles.