ERIA-1K: ERNIE-Image-Aes-1K, A Deployment-Oriented Image Aesthetics Benchmark with Realistic Distribution
[🤗 Dataset] [🧩 ERNIE-Image-Aes Model]
🔍 Overview
Existing aesthetic benchmarks are predominantly constructed from curated, high-production-value image collections, often sourced from platforms such as Flickr and DPChallenge. These datasets tend to skew toward professional or semi-professional photography communities, Western photographic traditions, and visually polished content, and therefore do not fully reflect the diversity of images encountered in real-world deployment.
We introduce ERIA-1K (ERNIE-Image-Aes-1K), an open-source human-annotated aesthetic benchmark designed to reflect realistic image distributions and provide a more deployment-oriented evaluation protocol.
🌟 Highlights
- 1,000 images spanning 6 categories with proportions calibrated to approximate real-world distribution
- Pairwise Swiss-system tournament annotation for stable and reproducible rankings
- Professional annotators from fine arts, design, and photography backgrounds
- Tier labels from 1 to 10 produced by calibrated annotators
- Designed to expose systematic biases of existing aesthetic predictors
📊 Benchmark Results
| Model | SRCC | PLCC |
|---|---|---|
| LAION AES | 0.2944 | 0.3138 |
| ArtiMuse | 0.4277 | 0.4704 |
| UniPercept | 0.4533 | 0.4748 |
| ERNIE-Image-Aes | 0.7445 | 0.7598 |
📂 Data Format
ERIA-1K-Benchmark/
└── test/
├── test.json
├── 000001.jpg
├── 000002.jpg
└── ...
- `images/`: contains the 1,000 benchmark images
- `test.json`: provides the ground-truth aesthetic tier scores for evaluation
## 📊 Evaluation
To evaluate your model on ERIA-1K, compute **SRCC** (Spearman's Rank Correlation Coefficient) and **PLCC** (Pearson's Linear Correlation Coefficient) between your model's predicted scores and the ground-truth tier labels in `test.json`.
## ✒️ Citation
If you find this benchmark useful, please consider citing:
```bibtex
@misc{ernie_image_aes_2025,
title={ERNIE-Image},
year={2025},
}