PEPL: Precision-Enhanced Pseudo-Labeling for Fine-Grained Image Classification in Semi-Supervised Learning

TL;DR

This work tackles the challenge of fine-grained image classification under semi-supervised learning by introducing Precision-Enhanced Pseudo-Labeling (PEPL), a two-stage framework that leverages Class Activation Maps (CAMs) to generate high-quality, semantic-aware pseudo-labels. Stage I performs adaptive confidence-thresholding with EMA updates to create provisional labels, while Stage II constructs Hybrid Semantic Pseudo-Labels through CAM-guided semantic mixing, producing labels that preserve fine-grained details. The approach is formalized with a combined loss, $L_{total} = \gamma L_{sup} + \lambda L_{unsup}$, where $L_{sup}$ and $L_{unsup}$ incorporate semantic proportions and CAM-based semantics. Experiments on CUB_200_2011 and Stanford Cars show PEPL achieving state-of-the-art accuracy under 10/20/30% labeled data, with semantic mixing yielding 4–9% gains in several settings, and qualitative visualizations reveal improved attention to discriminative fine-grained regions. The work demonstrates that CAM-guided pseudo-labeling and semantic mixing can substantially mitigate annotation costs in fine-grained SSL and provides open-source code for reproducibility.

Abstract

Fine-grained image classification has witnessed significant advancements with the advent of deep learning and computer vision technologies. However, the scarcity of detailed annotations remains a major challenge, especially in scenarios where obtaining high-quality labeled data is costly or time-consuming. To address this limitation, we introduce Precision-Enhanced Pseudo-Labeling(PEPL) approach specifically designed for fine-grained image classification within a semi-supervised learning framework. Our method leverages the abundance of unlabeled data by generating high-quality pseudo-labels that are progressively refined through two key phases: initial pseudo-label generation and semantic-mixed pseudo-label generation. These phases utilize Class Activation Maps (CAMs) to accurately estimate the semantic content and generate refined labels that capture the essential details necessary for fine-grained classification. By focusing on semantic-level information, our approach effectively addresses the limitations of standard data augmentation and image-mixing techniques in preserving critical fine-grained features. We achieve state-of-the-art performance on benchmark datasets, demonstrating significant improvements over existing semi-supervised strategies, with notable boosts in accuracy and robustness.Our code has been open sourced at https://github.com/TianSuya/SemiFG.

Figures (3)

• Figure 1: Instances where fine-grained details are corrupted by data augmentation
• Figure 2: The overview of our proposed methodology.
• Figure 3: Compared with method FreeMatch, the classifier obtained by method PEPL focuses more on fine-grained features such as car logos, lights, mirrors, and other more distinguishing parts.