docs: added colab line and fixed inle maths

README.md

@@ -1,6 +1,16 @@
+---
+license: apache-2.0
+---
+
 # Model Card for granite-geospatial-uki
 
-The granite-geospatial-uki model is a transformer-based geospatial foundation model trained on HLS L30 multispectral satellite imagery and Sentinel-1 synthetic aperture radar (SAR) backscatter over the United Kingdom and Ireland. The model consists of a self-supervised encoder developed with a ViT architecture and Masked AutoEncoder (MAE) learning strategy, with an MSE loss function and follows the same architecture as [Prithvi-EO](https://huggingface.co/collections/ibm-nasa-geospatial/prithvi-for-earth-observation-6740a7a81883466bf41d93d6).
+[<b><i>>>Try it on Colab<<</i></b>](https://colab.research.google.com/github/ibm-granite/geospatial/blob/main/uki-flooddetection/notebooks/2_fine_tuning.ipynb)
+
+The granite-geospatial-uki model is a transformer-based geospatial foundation model trained on HLS L30 multispectral satellite 
+imagery and Sentinel-1 synthetic aperture radar (SAR) backscatter over the United Kingdom and Ireland. 
+The model consists of a self-supervised encoder developed with a ViT architecture and Masked AutoEncoder (MAE) 
+learning strategy, with an MSE loss function and follows the same architecture as
+[Prithvi-EO](https://huggingface.co/collections/ibm-nasa-geospatial/prithvi-for-earth-observation-6740a7a81883466bf41d93d6).
 
 <figure>
     <img src='granite-geospatial-uki_image.png' alt='missing' />
@@ -15,7 +25,12 @@ An example of fine-tuning the model for image segmentation using Terratorch for
 
 ## Pre-training 
 
-The granite-geospatial-uki model uses the [Prithvi-EO](https://huggingface.co/collections/ibm-nasa-geospatial/prithvi-for-earth-observation-6740a7a81883466bf41d93d6) model architecture. It was pre-trained on HLS data for the continental USA, followed by additional pre-training using 16,000 HLS L30 and Sentinel-1 images covering the United Kingdom and Ireland. The Sentinel-1 SAR backscatter ($\sigma_0$) were resampled to the same resolution as the HLS data, and were normalized by $10log(\sigma_0)$, where pixels with $10log(\sigma_0) > 10$  are set to 10 and $10log(\sigma_0) < -35$ are set to -35. The two additional Sentinel-1 bands were initialized with the mean weights of the other channels for pre-training. The following bands were used in the pre-trained model:
+The granite-geospatial-uki model uses the [Prithvi-EO](https://huggingface.co/collections/ibm-nasa-geospatial/prithvi-for-earth-observation-6740a7a81883466bf41d93d6) 
+model architecture. It was pre-trained on HLS data for the continental USA, followed by additional pre-training using 16,000 HLS L30 and 
+Sentinel-1 images covering the United Kingdom and Ireland. The Sentinel-1 SAR backscatter (\\(\sigma_0\\)) were resampled to the same resolution
+as the HLS data, and were normalized by \\(10log(\sigma_0)\\), where pixels with \\(10log(\sigma_0) > 10\\)  are set to \\(10\\) and \\(10log(\sigma_0) < -35\\) 
+are set to \\(-35\\). The two additional Sentinel-1 bands were initialized with the mean weights of the other channels for pre-training.
+The following bands were used in the pre-trained model:
 
 - Blue
 - Green