FGDCC: Fine-Grained Deep Cluster Categorization -- A Framework for Intra-Class Variability Problems in Plant Classification
Published:Dec 24, 2025 05:00
•1 min read
•ArXiv AI
Analysis
This ArXiv paper introduces FGDCC, a novel method to address intra-class variability in Fine-Grained Visual Categorization (FGVC) tasks, specifically in plant classification. The core idea is to leverage classification performance by learning fine-grained features through class-wise cluster assignments. By clustering each class individually, the method aims to discover pseudo-labels that encode the degree of similarity between images, which are then used in a hierarchical classification process. While initial experiments on the PlantNet300k dataset show promising results and achieve state-of-the-art performance, the authors acknowledge that further optimization is needed to fully demonstrate the method's effectiveness. The availability of the code on GitHub facilitates reproducibility and further research in this area. The paper highlights the potential of cluster-based approaches for mitigating intra-class variability in FGVC.
Key Takeaways
- •FGDCC addresses intra-class variability in plant classification.
- •The method uses class-wise clustering to generate pseudo-labels.
- •Initial results on PlantNet300k are promising, but further optimization is needed.
Reference
“Our goal is to apply clustering over each class individually, which can allow to discover pseudo-labels that encodes a latent degree of similarity between images.”