ePrints@IIScePrints@IISc Home | About | Browse | Latest Additions | Advanced Search | Contact | Help

GrowBit: Incremental Hashing for Cross-Modal Retrieval

Mandal, D and Annadani, Y and Biswas, S (2019) GrowBit: Incremental Hashing for Cross-Modal Retrieval. In: 14th Asian Conference on Computer Vision, ACCV 2018, 2 - 6 December 2018, Perth, pp. 305-321.

[img] PDF
ACCV_2019.pdf - Published Version
Restricted to Registered users only

Download (476kB) | Request a copy
Official URL: https://doi.org/10.1007/978-3-030-20870-7_19


Cross-modal retrieval using hashing techniques is gaining increasing importance due to its efficient storage, scalability and fast query processing speeds. In this work, we address a related and relatively unexplored problem: given a set of cross-modal data with their already learned hash codes, can we increase the number of bits to better represent the data without relearning everything? This scenario is especially important when the number of tags describing the data increases, necessitating longer hash codes for better representation. To tackle this problem, we propose a novel approach called GrowBit, which incrementally learns the bits in the hash code and thus utilizes all the bits learned so far. We develop a two-stage approach for learning the hash codes and hash functions separately, utilizing a recent formulation which decouples over the bits so that it can incorporate the incremental approach. Experiments on MirFlickr, IAPR-TC-12 and NUS-WIDE datasets show the usefulness of the proposed approach.

Item Type: Conference Paper
Publication: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher: Springer Verlag
Additional Information: The copyright for this article belongs to Springer Verlag.
Keywords: Codes (symbols); Computer vision; Digital storage; Modal analysis, Cross-modal; Fast query; Hashing; Hashing techniques; Incremental approach; Incremental learning; Processing speed; Two stage approach, Hash functions
Department/Centre: Division of Electrical Sciences > Electrical Engineering
Date Deposited: 02 Dec 2022 09:56
Last Modified: 02 Dec 2022 09:56
URI: https://eprints.iisc.ac.in/id/eprint/77945

Actions (login required)

View Item View Item