Skip to content

🔒 Closed-set Classification

Overview

The classic supervised task: Given an image, assign it to one of the species present in the training set.
This is especially challenging in FungiTastic due to its fine-grained labels, long-tailed class distribution, and domain shifts.

Use Cases

  • Automated fungal species identification for field guides, citizen science apps, etc.
  • Baseline for comparing closed-world vision models.

Data & Splits

  • Uses the main FungiTastic split: training (all data up to end of 2021), validation (2022), test (2023).
  • Only species present in the training set are evaluated.

Evaluation Protocol

  • Primary Metric: Top-1 accuracy
  • Secondary Metrics: Macro F1-score, Top-3 accuracy

Baselines & Results

State-of-the-art models (e.g., ResNet, EfficientNet, ViT, BEiT) are provided as benchmarks. - Transformer architectures outperform classic CNNs, but all models struggle due to dataset complexity.

See Baselines & Models for scripts and results.

Quick Start