Update README.md

README.md

@@ -24,3 +24,17 @@ configs:
   - split: test
     path: data/test-*
 ---
+
+# 📄 BiblioPage
+
+**BiblioPage** dataset — a collection of 2,118 scanned title pages annotated with structured bibliographic metadata. The dataset serves as a benchmark for bibliographic metadata extraction, document understanding, and visual language model evaluation.
+
+## 🔍 About the Dataset
+
+- 2,118 annotated title pages from Czech libraries (1485–21st century)
+- 16 bibliographic attributes (e.g., Title, Author, Publication Date, etc.)
+- Annotations include both text and precise bounding boxes
+- Evaluation results available for YOLO, DETR, and VLLMs (GPT-4o, LLaMA 3)
+
+You can download the dataset on [Zenodo](https://zenodo.org/records/15683417).  
+For more details, see the [BiblioPage paper](https://arxiv.org/abs/2503.19658v1).