PROJECT TITLE :

Instance-Aware Hashing for Multi-Label Image Retrieval

ABSTRACT:

Similarity-preserving hashing is a commonly used method for nearest neighbor search in massive-scale image retrieval. For image retrieval, deep-network-primarily based hashing methods are appealing, since they'll simultaneously learn effective image representations and compact hash codes. This paper focuses on deep-network-primarily based hashing for multi-label images, each of which may contain objects of multiple classes. In most existing hashing strategies, each image is represented by one piece of hash code, that is known as semantic hashing. This setting might be suboptimal for multi-label image retrieval. To unravel this downside, we propose a deep architecture that learns instance-aware image representations for multi-label image data, that are organized in multiple teams, with every cluster containing the features for one class. The instance-aware representations not solely bring blessings to semantic hashing however conjointly will be utilized in class-aware hashing, in that an image is represented by multiple items of hash codes and each piece of code corresponds to a category. In depth evaluations conducted on many benchmark information sets demonstrate that for both the semantic hashing and also the class-aware hashing, the proposed technique shows substantial improvement over the state-of-the-art supervised and unsupervised hashing strategies.