Combining Discrepancy-Confusion Uncertainty and Calibration Diversity for Active Fine-Grained Image Classification
This work addresses the problem of efficient labeling in fine-grained image classification for researchers and practitioners, representing an incremental improvement over existing active learning methods.
The paper tackles the challenge of selecting informative samples for active learning in fine-grained image classification, where subtle inter-class differences make informativeness assessment difficult, and demonstrates that their method DECERN achieves superior performance compared to state-of-the-art methods across 7 datasets and 26 experimental settings.
Active learning (AL) aims to build high-quality labeled datasets by iteratively selecting the most informative samples from an unlabeled pool under limited annotation budgets. However, in fine-grained image classification, assessing this informativeness is especially challenging due to subtle inter-class differences. In this paper, we introduce a novel method, combining discrepancy-confusion uncertainty and calibration diversity for active fine-grained image classification (DECERN), to effectively perceive the distinctiveness between fine-grained images and evaluate the sample value. DECERN introduces a multifaceted informativeness measure that combines discrepancy-confusion uncertainty and calibration diversity. The discrepancy-confusion uncertainty quantifies the category directionality and structural stability of fine-grained unlabeled data during local feature fusion. Subsequently, uncertainty-weighted clustering is performed to diversify the uncertainty samples. Then we calibrate the diversity to maximize the global diversity of the selected sample while maintaining its local representativeness. Extensive experiments conducted on 7 fine-grained image datasets across 26 distinct experimental settings demonstrate that our method achieves superior performance compared to state-of-the-art methods.