update readme file

README.md

@@ -1,36 +1,86 @@
 ---
 license: mit
+language:
+- si
 base_model: kavg/TrOCR-SIN-DeiT
+pipeline_tag: image-to-text
 tags:
-- generated_from_trainer
-model-index:
-- name: TrOCR-SIN-DeiT-Handwritten
-  results: []
+- ocr
+- handwriting
+- sinhala
+- trocr
 ---
 
-<!-- This model card has been generated automatically according to the information the Trainer had access to. You
-should probably proofread and complete it, then remove this comment. -->
+# TrOCR-Sin-Handwritten
 
-# TrOCR-SIN-DeiT-Handwritten
+A Sinhala text-line recognizer for **handwritten** text, released with the SinFUND and SinOCR datasets ([SriDoc/datasets](https://github.com/SriDoc/datasets)).
 
-This model is a fine-tuned version of [kavg/TrOCR-SIN-DeiT](https://huggingface.co/kavg/TrOCR-SIN-DeiT) on an unknown dataset.
-It achieves the following results on the evaluation set:
-- Loss: 2.9839
-- Cer: 0.5253
+## Model
 
-## Model description
+- TrOCR-style vision encoder-decoder (`VisionEncoderDecoderModel`).
+- The encoder was initialized from [`facebook/deit-base-distilled-patch16-224`](https://huggingface.co/facebook/deit-base-distilled-patch16-224) (ImageNet weights). The decoder was initialized from [`NLPC-UOM/SinBERT-large`](https://huggingface.co/NLPC-UOM/SinBERT-large), a RoBERTa model pre-trained on Sinhala text. The cross-attention layers started from random weights.
+- No OCR-pretrained TrOCR weights were used.
+- **Input:** one pre-cropped text-line image, resized to 224 x 224.
+- **Decoding:** beam search with 4 beams and a maximum length of 64 tokens.
 
-More information needed
+## Training data
 
-## Intended uses & limitations
+This model was trained in two stages:
+1. On the SinOCR printed training split. That model is [SriDoc/TrOCR-Sin-Printed](https://huggingface.co/SriDoc/TrOCR-Sin-Printed).
+2. Then on the SinOCR handwritten training split (908 lines) for 100 epochs (2,900 steps), with a batch size of 32 and a learning rate of 5e-5.
 
-More information needed
+## Results
 
-## Training and evaluation data
+Character error rate (CER; lower is better) on the SinOCR test splits:
 
-More information needed
+| Test split | CER |
+|---|---|
+| SinOCR handwritten (227 lines) | 0.5253 |
+| SinOCR printed (10,000 lines) | 0.2246 |
 
-## Training procedure
+How the CER is computed:
+- Characters are Unicode code points.
+- The CER is pooled over the whole test split.
+
+SinOCR has no separate validation split, so the test split was also the evaluation set during training. The lowest test CER occurred at the final checkpoint, so checkpoint selection did not change this result.
+
+## Limitations
+
+- **Errors remain common.** A CER of about 0.53 means roughly one character error for every two characters, so the output needs human checking.
+- **Test writers were seen in training.** The handwritten test split is not writer-disjoint: 72% of its lines come from writers who also appear in training. The CER is therefore probably optimistic for new writers.
+- **Do not use this model to measure OCR accuracy on the SinFUND test forms.** 132 of its training lines are crops of those forms. The dataset repository lists them in `sinocr_handwritten_provenance.csv`.
+- The model reads single, pre-cropped lines. It does not find text on a page.
+
+## Usage
+
+```python
+from transformers import VisionEncoderDecoderModel, TrOCRProcessor, AutoTokenizer, ViTImageProcessor
+from PIL import Image
+
+model = VisionEncoderDecoderModel.from_pretrained("SriDoc/TrOCR-Sin-Handwritten")
+tokenizer = AutoTokenizer.from_pretrained("NLPC-UOM/SinBERT-large")
+image_processor = ViTImageProcessor.from_pretrained("google/vit-base-patch16-224")
+processor = TrOCRProcessor(image_processor=image_processor, tokenizer=tokenizer)
+
+image = Image.open("line.png").convert("RGB")   # one cropped text line
+pixel_values = processor(image, return_tensors="pt").pixel_values
+generated_ids = model.generate(pixel_values)
+print(processor.batch_decode(generated_ids, skip_special_tokens=True)[0])
+```
+
+## Citation
+
+```bibtex
+@article{gunathilaka2025sinfund,
+  title   = {SinFUND and SinOCR: Benchmarks for Sinhala Handwritten OCR and Template-Free Form Understanding},
+  author  = {Gunathilaka, Kavishka and Hewagama, Danusha and Pushpakumara, Supul and Ambegoda, Thanuja D.},
+  journal = {Research Square preprint},
+  year    = {2025},
+  doi     = {10.21203/rs.3.rs-6976719/v1}
+}
+```
+
+## Training log
 
 ### Training hyperparameters
 
@@ -46,6 +96,8 @@ The following hyperparameters were used during training:
 
 ### Training results
 
+The "Cer" and "Validation Loss" columns were measured on the SinOCR handwritten **test** split.
+
 | Training Loss | Epoch | Step | Validation Loss | Cer    |
 |:-------------:|:-----:|:----:|:---------------:|:------:|
 | 0.2915        | 3.45  | 100  | 1.8613          | 0.6450 |
@@ -78,7 +130,6 @@ The following hyperparameters were used during training:
 | 0.0           | 96.55 | 2800 | 3.1309          | 0.5273 |
 | 0.0           | 100.0 | 2900 | 2.9839          | 0.5253 |
 
-
 ### Framework versions
 
 - Transformers 4.35.2