---
pretty_name: FG-BMK Image Archive - iNat2021 validation
license: other
language:
- en
task_categories:
- image-classification
tags:
- fine-grained-recognition
- large-vision-language-models
- benchmark
- image-dataset
---

# FG-BMK Image Archive: iNat2021 validation

This repository stores the iNat2021 validation image archive used by FG-BMK. It is a companion repository for [SEU-VIPGroup/FG-BMK](https://huggingface.co/datasets/SEU-VIPGroup/FG-BMK), which contains the benchmark metadata, questions, splits, and evaluation code.

## Files

- `val.tar.gz`: image archive for the iNat2021 validation subset.

## Use with FG-BMK

1. Download `val.tar.gz`.
2. Extract it to a local data directory.
3. Use [SEU-VIPGroup/FG-BMK](https://huggingface.co/datasets/SEU-VIPGroup/FG-BMK) for the corresponding question files, class lists, train/test splits, and evaluation scripts.
4. Point the FG-BMK evaluation scripts to the extracted image folder.

## Citation

```bibtex
@article{yu2026fgbmk,
  title   = {Benchmarking Large Vision-Language Models on Fine-Grained Image Tasks: From Evaluation to Diagnosis},
  author  = {Yu, Hong-Tao and Xie, Chen-Wei and Peng, Yuxin and Belongie, Serge and Wei, Xiu-Shen},
  journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)},
  year    = {2026}
}
```

## License and Data Terms

This archive is provided for FG-BMK evaluation. Users must follow the license and usage terms of the original iNat2021 validation dataset as well as the FG-BMK project terms.