---
license: cc-by-nc-sa-4.0
task_categories:
- image-text-to-image
language:
- en
pretty_name: icir
size_categories:
- 100K<n<1M
---
## i-CIR Dataset (Hugging Face)

[**website**](https://vrg.fel.cvut.cz/icir/)  |  [**arxiv**](https://arxiv.org/pdf/2510.25387) |  [**github**](https://github.com/billpsomas/icir)

### About
**i-CIR (Instance-Level Composed Image Retrieval)** is a curated benchmark for **composed image retrieval** where each *instance* corresponds to a specific, visually indistinguishable object (e.g., a particular landmark). Each query combines an **image of the instance** with a **text modification**, and retrieval is evaluated against a database containing **rich hard negatives** (visual / textual / compositional).

<p align="center">
<img width="75%" alt="i-CIR illustration" src="https://github.com/billpsomas/icir/raw/main/.github/dataset.png">
</p>

**Key stats**
- **Instances:** 202  
- **Total images:** ~750K  
- **Composed queries:** 1,883  
- **Avg database size / query:** ~3.7K images  
- Includes challenging hard negatives per instance.

---

### Dataset Structure
On Hugging Face, i-CIR is hosted as **WebDataset shards** for scalable/robust downloads and streaming.

```text
icir/
├── webdataset/
│   ├── query/
│   │   ├── query-000000.tar
│   │   ├── query-000001.tar
│   │   └── ...
│   └── database/
│       ├── database-000000.tar
│       ├── database-000001.tar
│       └── ...
├── annotations/
│   ├── query_files.csv
│   ├── database_files.csv
├── VERSION.txt
└── LICENSE
```

---

### Annotations format

- query_files.csv: each row is (image_path, text_query, instance_id)
- database_files.csv: each row is (image_path, text_query, instance_id) (the text field may be unused for database features depending on the pipeline)

Inside each WebDataset sample, we store:
- an image (.jpg/.png/...)
- a json payload with: img_path, text, instance

---

### Download

One-liner download (recommended):

```bash
pip install -U huggingface_hub
huggingface-cli download billpsomas/icir --repo-type dataset --local-dir ./data/icir --revision main
```

Python (equivalent):

```python
from huggingface_hub import snapshot_download
snapshot_download(repo_id="billpsomas/icir", repo_type="dataset", local_dir="./data/icir", revision="main")
```

---

### Using the dataset (feature extraction)

You can extract features directly from the WebDataset shards (no image folder extraction needed):

```bash
python3 create_features.py \
  --dataset icir \
  --icir_source wds \
  --icir_wds_root ./data/icir \
  --backbone clip \
  --batch 512 \
  --gpu 0
```

---

### License

The dataset is released under CC BY-NC-SA 4.0. Please see LICENSE for details.

---

### Citation

If you use i-CIR in your research, please cite:

```
@inproceedings{
    psomas2025instancelevel,
    title={Instance-Level Composed Image Retrieval},
    author={Bill Psomas and George Retsinas and Nikos Efthymiadis and Panagiotis Filntisis and Yannis Avrithis and Petros Maragos and Ondrej Chum and Giorgos Tolias},
    booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems},
    year={2025}
}
```