update readme

README.md

@@ -11,4 +11,24 @@ pinned: false
 short_description: 'Trying out different binarizations solutions '
 ---
 
+# Xibi Binarization
+
+This demo compares multiple document binarization methods:
+- `Sauvola (Traditional)` — CPU-based adaptive thresholding
+- `Tzefa-Binarization` — Hugging Face image-to-image pipeline fallback
+- `Two-Stage GAN (opensuh)` — Hugging Face pipeline fallback
+- `DocRes (Generalist Transformer)` — Hugging Face pipeline fallback
+
+## Notes
+
+- `app.py` now supports running even when `cv2` is not installed.
+- If the hosted Hugging Face model pipelines are unavailable, the demo falls back to local grayscale thresholding or Sauvola-like binarization.
+- This makes the app more stable in environments where GPU/model dependencies are missing.
+
+## Requirements
+
+Install the standard dependencies listed in `requirements.txt`.
+
+If you want full model support, also install and configure any required GPU/transformers support for the Hugging Face pipelines.
+
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference