🧪 Baselines & Models Overview
FungiTastic includes strong, diverse baselines for all benchmark tasks.
You’ll find:
- Ready-to-use pre-trained weights and scripts
- Classic CNNs, modern transformers, and multimodal models
- Results on all main benchmarks
📋 Model Categories
| Task | Model Page | Baseline Types |
|---|---|---|
| Closed-set Classification | Closed-set Models | CNN, ViT, BEiT, etc. |
| Few-shot Learning | Few-shot Models | ProtoNet, NN, CLIP, etc. |
| Open-set Classification | Open-set Models | Softmax, logit, DINOv2 |
| Vision-Language Fusion | VLM Fusion | DistilBERT, Fusion |
Segmentation models are covered in the Segmentation Baselines.
🚀 Getting Started
- All model scripts/notebooks are in the GitHub repo
- Download checkpoints and splits from Kaggle
- Benchmark protocols: see usage/training.md and usage/evaluation.md
Choose a model page for more details on architectures, training, and results.