image-retrieval-models

Deep Orthogonal Fusion of Local and Global Features

The topic of Deep Orthogonal Local and Global (DOLG) information fusion framework for generating image representations is aimed at developing an effective single-stage solution for image retrieval by integrating local and global information within images. The aim of image retrieval is to obtain images similar to a query image from a database, with a common practice of retrieving candidate images through similarity searches using global features, and then re-rank the choices by leveraging their l

DELG

DELG is a powerful neural network designed for image retrieval using a combination of techniques for global and local features. This innovative model can be trained end-to-end, requiring only image-level labeling, and is optimized to extract an image’s global feature, detect keypoints, and create local descriptors all within a single model. How DELG Works At its core, DELG utilizes hierarchical image representations that are produced by convolutional neural networks (CNNs), which are then pai