Comparison of Pairwise Similarity Distance Methods for Effective Hashing

Abstract Content-based image retrieval (CBIR) methods search for points with the most similar content to query features from within a large dataset. The notable approach this purpose is an approximate nearest neighbor (ANN) searching. main properties expected system can be listed as follows; low storage requirement, high retrieve speed, and average precision. Hashing, which generate discriminative low-dimensional binary codes, one of today’s effective ANN searching methods. Although there are various hashing approaches in literature, almost all consist low-dimension feature representation binarization sections. This study focuses on representation. Hand-crafted or deep learning based used extraction These components that affect performance creating codes. Contrastive loss often literature update learnable parameters these algorithms. distance parameter data critical calculating contrastive loss. In study, tested using five different (Euclidean, Manhattan, Cosine, Minkowski, Chebyshev) more Retrieval vectors produced by MNIST CIFAR-10 datasets. It thought information obtained very useful new researchers.