Transfer Learning and Mixup for Fine-Grained Few-Shot Fungi Classification

TL;DR

This paper explores transfer learning, data augmentation, and multimodal strategies for fine-grained fungi classification, achieving competitive results in a challenging few-shot visual categorization task.

Contribution

It introduces a comprehensive approach combining vision transformers, textual data, and sampling techniques for improved fungi species identification in few-shot settings.

Findings

Vision transformer models outperform traditional CNNs.

Domain-specific pretraining enhances classification accuracy.

Balanced sampling strategies improve model robustness.

Abstract

Accurate identification of fungi species presents a unique challenge in computer vision due to fine-grained inter-species variation and high intra-species variation. This paper presents our approach for the FungiCLEF 2025 competition, which focuses on few-shot fine-grained visual categorization (FGVC) using the FungiTastic Few-Shot dataset. Our team (DS@GT) experimented with multiple vision transformer models, data augmentation, weighted sampling, and incorporating textual information. We also explored generative AI models for zero-shot classification using structured prompting but found them to significantly underperform relative to vision-based models. Our final model outperformed both competition baselines and highlighted the effectiveness of domain specific pretraining and balanced sampling strategies. Our approach ranked 35/74 on the private test set in post-completion evaluation, this suggests additional work can be done on metadata selection and domain-adapted multi-modal learning. Our code is available at https://github.com/dsgt-arc/fungiclef-2025.