Dataset card

README.md

@@ -1,21 +1,45 @@
 ---
-dataset_info:
-  features:
-  - name: image
-    dtype: image
-  - name: label
-    dtype: string
-  - name: file_name
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 9950864
-    num_examples: 5000
-  download_size: 9902504
-  dataset_size: 9950864
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
+license: mit
+language:
+- km
+task_categories:
+- image-to-text
+tags:
+- ocr
+- khmer-ocr
+- benchmark
+- synthetic
+pretty_name: Khmer Printed Words
+size_categories:
+- 1K<n<10K
 ---
+
+# Khmer Printed Words
+
+5,000 synthetic images of short, isolated Khmer words in 10 fonts, with their text. Most training lines are mid-to-long, so this set tests short input.
+
+It is the "Printed Words" test set in the [Netra OCR](https://github.com/netra-ai-lab/Netra-OCR) results, which use its first 500 lines.
+
+| Column | |
+| --- | --- |
+| `image` | the word image |
+| `label` | its text |
+| `file_name` | the original file name |
+
+## Use
+
+```python
+from datasets import load_dataset
+ds = load_dataset("Netra-Lab/khmer-printed-words", split="train")
+```
+
+With the Netra OCR repository, the `printed-words` preset scores any model on it:
+
+```bash
+bash training/evaluate.sh published --test printed-words
+bash training/evaluate.sh published --test Netra-Lab/khmer-printed-words:train[:500]   # as in the results
+```
+
+## License
+
+MIT. Made by Netra Lab.