README.md

---
configs:
- config_name: Absolute_depth
  data_files:
  - split: train
    path: train/Absolute_depth/train-*.parquet
- config_name: Action
  data_files:
  - split: train
    path: train/Action/train-*.parquet
- config_name: Color
  data_files:
  - split: train
    path: train/Color/train-*.parquet
- config_name: Counting
  data_files:
  - split: train
    path: train/Counting/train-*.parquet
- config_name: Emotion
  data_files:
  - split: train
    path: train/Emotion/train-*.parquet
- config_name: Fine-grained
  data_files:
  - split: train
    path: train/Fine-grained/train-*.parquet
- config_name: Localization
  data_files:
  - split: train
    path: train/Localization/train-*.parquet
- config_name: OCR
  data_files:
  - split: train
    path: train/OCR/train-*.parquet
- config_name: Orientation
  data_files:
  - split: train
    path: train/Orientation/train-*.parquet
- config_name: Recognition
  data_files:
  - split: train
    path: train/Recognition/train-*.parquet
- config_name: Relative_depth
  data_files:
  - split: train
    path: train/Relative_depth/train-*.parquet
- config_name: Scene_Classification
  data_files:
  - split: train
    path: train/Scene_Classification/train-*.parquet
- config_name: Spatial
  data_files:
  - split: train
    path: train/Spatial/train-*.parquet
- config_name: Texture
  data_files:
  - split: train
    path: train/Texture/train-*.parquet
license: cc-by-4.0
task_categories:
- visual-question-answering
- image-classification
- image-to-text
language:
- en
tags:
- multimodal
- vision-language
- benchmark
- instruction-tuning
size_categories:
- 100K<n<1M
---

# AVA-Bench

Training dataset for the paper **AVA-Bench: Atomic Visual Ability Benchmark for Vision Foundation Models** ([arXiv:2506.09082](https://arxiv.org/abs/2506.09082)) accepted in **CVPR 2026**.

AVA-Bench is a diagnostic benchmark for evaluating Vision Foundation Models (VFMs) through Atomic Visual Abilities (AVAs): fundamental perceptual skills such as localization, counting, OCR, spatial understanding, depth estimation, color recognition, texture recognition, and fine-grained recognition. 

AVA-Bench disentangls visual perception into **14 atomic visual capabilities**, each with distribution-matched training and evaluation splits. This allows researchers to measure where a VFM excels or fails and to construct capability-level “ability fingerprints” for model comparison and selection. 

This Hub release contains the **training split** of AVA-Bench. The evaluation split is released separately; see the project page and paper for details.

## Capabilities

AVA-Bench covers **14 atomic visual capabilities**, each released as its own subset/config:

| Capability | Tests |
|---|---|
| `Action` | Recognizing human/animal actions in images |
| `Color` | Identifying object colors |
| `Counting` | Counting instances of an object |
| `Emotion` | Recognizing emotion from facial expressions/scenes |
| `Fine-grained` | Fine-grained category discrimination, such as bird, plant, animal, fungi, or aircraft categories |
| `Localization` | Locating objects via bounding-box queries |
| `OCR` | Reading text rendered in images |
| `Orientation` | Determining the orientation or pose of objects |
| `Recognition` | Object/entity recognition |
| `Scene_Classification` | Classifying the overall scene/place |
| `Spatial` | Reasoning about spatial relationships between objects |
| `Texture` | Identifying surface textures |
| `Absolute_depth` | Estimating absolute depth from a single image |
| `Relative_depth` | Comparing depth between two regions |

## Dataset structure

Each subset has a single `train` split, stored as Parquet shards with image bytes **embedded** in the file.

### Data fields

Every example contains:

- `image` (`datasets.Image`) — the input image, decoded as a PIL image on access.
- `id` (`string`) — unique example identifier.
- `conversations` (`list` of `{from, value}`) — instruction-tuning style turns. The `human` turn includes the question, usually with an `<image>` placeholder, and the `gpt` turn includes the ground-truth answer.

Some capabilities may additionally include fields such as:

- `height`
- `width`
- `category`
- `area`
- bounding-box or region metadata, depending on the capability
Per-subset row counts are visible in the dataset viewer's config dropdown.

## Usage
Please go to github to use the dataset to evaluate Vision Foundation Models. If you want to check the dataset:

```python
from datasets import load_dataset

# Load one capability
ds = load_dataset("act13/AVA-Bench", name="Counting", split="train")

print(ds[0])
# {
#   'id': '...',
#   'image': <PIL.Image.Image image mode=RGB ...>,
#   'conversations': [
#       {'from': 'human', 'value': '<image>\n...'},
#       {'from': 'gpt', 'value': '...'}
#   ],
#   ...
# }
```

To stream without downloading the full subset:

```python
from datasets import load_dataset

ds = load_dataset(
    "act13/AVA-Bench",
    name="Counting",
    split="train",
    streaming=True,
)

for ex in ds.take(5):
    print(ex["conversations"][0]["value"])
    print(ex["conversations"][1]["value"])
```

## Intended uses

AVA-Bench is intended for research on vision foundation models and vision-language systems. Suitable uses include:

- Training or instruction-tuning vision-language models on atomic visual abilities.
- Diagnosing which visual capabilities a VFM lacks.
- Comparing VFMs through capability-level performance rather than only aggregate VQA accuracy.
- Constructing balanced training mixtures across visual abilities.
- Studying how different VFM pretraining objectives affect downstream perceptual capabilities.



## Source datasets

AVA-Bench is curated from multiple existing datasets, depending on the atomic visual ability. Source datasets include, but are not necessarily limited to:

- Objects365
- LVIS
- iNaturalist-2021
- DIOR
- NYU-Depth V2
- KITTI
- COCO-Text
- IIIT5K
- TextVQA
- EgoOrientBench
- CURE-OR
- Places434
- AID
- CUB-200-2011
- FGVC-Aircraft
- MiT
- DTD
- Kylberg
- KTH-TIPS
- KTH-TIPS2

Please see the paper for the full per-capability dataset construction details and source-license breakdown.

## License

This dataset card and the AVA-Bench organization/annotations are released under **CC-BY-4.0**.

Underlying images retain the licenses of their original source datasets. Users are responsible for respecting the license terms and usage restrictions of each source dataset.

## Citation

If you use AVA-Bench, please cite:

```bibtex
@article{mai2025ava,
  title={Ava-bench: Atomic visual ability benchmark for vision foundation models},
  author={Mai, Zheda and Chowdhury, Arpita and Wang, Zihe and Jeon, Sooyoung and Wang, Lemeng and Hou, Jiacheng and Chao, Wei-Lun},
  journal={arXiv preprint arXiv:2506.09082},
  year={2025}
}
```

## Contact

Open a discussion on this dataset's Community tab, or reach the authors via the contact information provided in the paper.