Update README for the full version of the dataset

README.md

@@ -16,60 +16,60 @@ size_categories:
 ---
 
 
-# Dataset for Evaluating the Aesthetics of Rendering (DEAR) - 100-scene Sample
+# REPID: Rendering Evaluation of Photographic Image Dataset
 
-This dataset is a preliminary release of the **Dataset for Evaluating the Aesthetics of Rendering (DEAR)**, designed to model human aesthetic judgments of image rendering styles. Built upon the MIT-Adobe FiveK dataset, DEAR provides a foundation for training and evaluating models that can assess subjective rendering preferences rather than just technical image quality.
+ **REPID** (officially introduced as the Rendering Evaluation of Photographic Image Dataset) is a large-scale benchmark designed for **Image Rendering Quality Assessment (IRQA)**.
+> Unlike traditional Image Quality Assessment (IQA) which focuses on technical degradations like noise or blur, REPID aims to model subjective human aesthetic preferences for different rendering styles of the same scene.
 
 ## Dataset Overview
 
-This sample contains **100 scenes** (600 total images) with **pairwise human preference annotations** collected through large-scale crowdsourcing. Each scene is rendered in **6 distinct styles**:
+> Built upon the MIT-Adobe FiveK dataset, REPID provides a massive collection of pairwise human preference annotations for professional and automated renderings.
 
-- **Style A-E**: Five different photographer-specific renderings (Expert A through Expert E from the original MIT-Adobe FiveK dataset)
-- **Neutral**: Adobe Photoshop auto mode rendering (serves as a baseline/reference style)
+> * **Scenes**: 5,000 high-resolution RAW photographs.
+> * **Total Images**: 30,000 unique renderings (6 per scene).
+> * **Total Votes**: Over 2.5 million unique votes collected via crowdsourcing.
+> * **Annotators**: 13,648 unique evaluators, with each image pair receiving at least 25 individual votes.
+> * **Comparison Task**: For each of the 15 possible pairs per scene, evaluators indicated "Left preferable", "Right preferable", or "Both equal".
 
-Each pair of renderings for the same scene was evaluated by multiple human annotators who indicated which version they preferred or if they considered them equally appealing. This rich preference data enables training models to predict aesthetic preferences across different rendering styles.
-
-|style a|style b|style c|style d|style e|neutral|
-|-------|-------|-------|-------|-------|-------|
-|<img src="./images/tiff16_a/a2031-WP_CRW_5715.jpeg">|<img src="./images/tiff16_b/a2031-WP_CRW_5715.jpeg">|<img src="./images/tiff16_c/a2031-WP_CRW_5715.jpeg">|<img src="./images/tiff16_d/a2031-WP_CRW_5715.jpeg">|<img src="./images/tiff16_e/a2031-WP_CRW_5715.jpeg">|<img src="./images/original/a2031-WP_CRW_5715.jpeg">|
-|<img src="./images/tiff16_a/a3214-KE_-8375.jpeg">|<img src="./images/tiff16_b/a3214-KE_-8375.jpeg">|<img src="./images/tiff16_c/a3214-KE_-8375.jpeg">|<img src="./images/tiff16_d/a3214-KE_-8375.jpeg">|<img src="./images/tiff16_e/a3214-KE_-8375.jpeg">|<img src="./images/original/a3214-KE_-8375.jpeg">|
-|<img src="./images/tiff16_a/a0196-2004-01-25 13-34-10 CRW_3079.jpeg">|<img src="./images/tiff16_b/a0196-2004-01-25 13-34-10 CRW_3079.jpeg">|<img src="./images/tiff16_c/a0196-2004-01-25 13-34-10 CRW_3079.jpeg">|<img src="./images/tiff16_d/a0196-2004-01-25 13-34-10 CRW_3079.jpeg">|<img src="./images/tiff16_e/a0196-2004-01-25 13-34-10 CRW_3079.jpeg">|<img src="./images/original/a0196-2004-01-25 13-34-10 CRW_3079.jpeg">|
-|<img src="./images/tiff16_a/a0036-jn_2007_05_05__183.jpeg">|<img src="./images/tiff16_b/a0036-jn_2007_05_05__183.jpeg">|<img src="./images/tiff16_c/a0036-jn_2007_05_05__183.jpeg">|<img src="./images/tiff16_d/a0036-jn_2007_05_05__183.jpeg">|<img src="./images/tiff16_e/a0036-jn_2007_05_05__183.jpeg">|<img src="./images/original/a0036-jn_2007_05_05__183.jpeg">|
-|<img src="./images/tiff16_a/a3308-Ja_Pe-23.jpeg">|<img src="./images/tiff16_b/a3308-Ja_Pe-23.jpeg">|<img src="./images/tiff16_c/a3308-Ja_Pe-23.jpeg">|<img src="./images/tiff16_d/a3308-Ja_Pe-23.jpeg">|<img src="./images/tiff16_e/a3308-Ja_Pe-23.jpeg">|<img src="./images/original/a3308-Ja_Pe-23.jpeg">|
-|<img src="./images/tiff16_a/a0742-IMG_2429.jpeg">|<img src="./images/tiff16_b/a0742-IMG_2429.jpeg">|<img src="./images/tiff16_c/a0742-IMG_2429.jpeg">|<img src="./images/tiff16_d/a0742-IMG_2429.jpeg">|<img src="./images/tiff16_e/a0742-IMG_2429.jpeg">|<img src="./images/original/a0742-IMG_2429.jpeg">|
+### Rendering Styles
+Each scene features six distinct interpretations:
+* **Expert A–E**: Five professional photographer renderings from the original MIT-Adobe FiveK dataset.
+* **Neutral**: Adobe Photoshop’s "Auto" mode rendering, serving as a baseline.
 
 ## Key Features
 
-- **Human-centered evaluation**: Captures nuanced aesthetic preferences rather than technical distortions
-- **Multiple rendering styles**: Six distinct interpretations per scene allow for fine-grained preference analysis
-- **Expert vs. automated rendering**: Comparison between professional photographer styles and Adobe's auto mode
-- **Diverse content**: Scenes span portraits, landscapes, food, animals, night photography, and general vernacular subjects
+* **Subjective Focus**: Targets the "aesthetics of rendering" (color, texture, and artistic expression) rather than simple signal-to-noise ratios.
 
-## Metadata
+* **Content-Dependent Preferences**: The dataset reveals that preferred rendering styles vary significantly based on image content (e.g., portraits vs. landscapes).
 
-Each image includes comprehensive metadata extracted from the original TIFF XMP data, including:
+* **Personalization Support**: Includes unique evaluator identifiers, enabling research into personalized aesthetic preference modeling—a critical area for recommendation systems and generative AI[cite: 1622].
+* **Content Annotations**: Includes scene classifications generated via BLIP (e.g., nature, food, night scenes) with verified 96% accuracy.
 
-- **Technical parameters**: Exposure settings, white balance, contrast, saturation, and tone curve adjustments
-- **Rendering information**: Software version, processing history, and renderer identifiers
-- **Content categories**: Auto-generated scene classifications (people, animals, food, night scenes, etc.)
-- **Image properties**: Dimensions, color profiles, and format specifications
-- **Source information**: Original filename, capture device, and photographer attribution
+## Dataset Statistics
 
-The metadata provides valuable context for understanding how specific parameter adjustments influence aesthetic preferences and enables researchers to analyze correlations between technical settings and human preferences.
+| Feature | Value |
+| :--- | :--- |
+| **Scenes** | 5,000  |
+| **Images** | 30,000  |
+| **Annotators** | 13,648  |
+| **Votes** | 2,500,000+  |
+| **Test Set** | 1,283 scenes (approx. 25% of data)  |
+| **Upper Bound Accuracy** | 0.896 (human consensus)  |
 
 ## Applications
 
-This dataset enables research in:
-- **Aesthetic preference prediction**: Training models to predict which rendering style users will prefer
-- **Style-aware image enhancement**: Developing algorithms that adapt to user aesthetic preferences
-- **Personalized rendering**: Creating systems that learn individual aesthetic tastes
-- **Benchmarking aesthetic models**: Evaluating existing and new approaches to aesthetic quality assessment
-- **Human-AI collaboration**: Understanding how human preferences can guide AI rendering decisions
+REPID is designed to foster research in:
+
+1.  **Aesthetic Preference Prediction**: Training models to predict which of two renderings a human will prefer.
+2.  **Personalized Rendering**: Modeling individual user tastes using the provided evaluator IDs.
+3.  **Render Ranking**: Developing systems that can automatically select the "best" rendering for a given image.
+4.  **Benchmarking IRQA**: Providing a qualitatively different challenge than traditional distortion-based IQA benchmarks.
+
 
 ## Citation
 ```
 @misc{plohotnyuk2025dear,
-  title={DEAR: Dataset for Evaluating the Aesthetics of Rendering},
+  title={Beyond distortions — a benchmark for subjective evaluation of image rendering quality},
   author={Plohotnyuk, Vsevolod and Panshin, Artyom and Bani{\'c}, Nikola and Bianco, Simone and Freeman, Michael and Ershov, Egor},
   year={2025},
   eprint={2512.05209},