Weakly Supervised Attention-based Recognition under Spectral, Turbulence, and Resource Variations

Authors

Kshitij Naresh Nikhal, University of Nebraska-LincolnFollow

First Advisor

Benjamin Riggan

Department

Electrical Engineering

Date of this Version

Spring 2024

Document Type

Dissertation

Comments

Copyright 2024, Kshitij Naresh Nikhal. Used by permission

Abstract

While supervised optimization paradigms are ubiquitous across diverse recognition systems, the risk of over-fitting and increasing bias have limited their applicability. This dissertation focuses on unsupervised learning—learning without precisely curated data—and argues that unsupervised learning methods can enable both discriminability and generalizability. Through the use of attention-based machine learning and advanced clustering, unsupervised methods are able to focus on fine-grained information in images without any explicit supervision. The dissertation introduces a domain-bridging framework for tasks like cross-spectrum matching and long-range recognition, utilizing intra-domain clustering and inter-domain matching to generate pseudo-labels. Additionally, a hash-based network is proposed to accelerate the search process, with dynamic capabilities for computational adjustment based on operational demands or input complexity. The overall framework is demonstrated over biometric tasks, such as person re-identification and facial recognition, and shows superiority in performance and efficiency compared to other unsupervised methods, and in some cases even supervised methods.

Recommended Citation

Nikhal, Kshitij Naresh, "Weakly Supervised Attention-based Recognition under Spectral, Turbulence, and Resource Variations" (2024). Dissertations and Doctoral Documents from University of Nebraska-Lincoln, 2023–. 56.
https://digitalcommons.unl.edu/dissunl/56