RAUM-Net: Regional Attention and Uncertainty-aware Mamba Network

TL;DR

RAUM-Net is a semi-supervised approach for fine-grained visual categorization that combines regional attention, Bayesian uncertainty, and Mamba-based feature modeling to improve robustness with limited labeled data.

Contribution

It introduces a novel semi-supervised method integrating regional attention, Bayesian uncertainty, and Mamba networks for enhanced FGVC performance.

Findings

Strong performance on FGVC benchmarks with occlusions.

Robustness when labeled data is limited.

Effective pseudo label selection via Bayesian inference.

Abstract

Fine Grained Visual Categorization (FGVC) remains a challenging task in computer vision due to subtle inter class differences and fragile feature representations. Existing methods struggle in fine grained scenarios, especially when labeled data is scarce. We propose a semi supervised method combining Mamba based feature modeling, region attention, and Bayesian uncertainty. Our approach enhances local to global feature modeling while focusing on key areas during learning. Bayesian inference selects high quality pseudo labels for stability. Experiments show strong performance on FGVC benchmarks with occlusions, demonstrating robustness when labeled data is limited. Code is available at https://github.com/wxqnl/RAUM Net.