completing docs cameroon example

docs/source/examples.md

@@ -5,6 +5,8 @@
 
 The goal in this example is to map different forest types using multispectral Sentinel 2 data.
 
+We have a raster and ... points that have been labelled by photo-interpretation.
+
 ```{image} ./_static/examples/original_points.png
 :alt: Original image with GT points
 :class: centered-image
@@ -12,13 +14,26 @@ The goal in this example is to map different forest types using multispectral Se
 :align: center
 ```
 
-Example table (temp)
+Here is a recap of the random forest accuracy given different pre- and post-processing of the encoder's features.
+
+| Using a DL encoder    | Pre-processing    | Post-processing   | Accuracy  |
+| :-----------          | :-----------      | :--------------   |---------: |
+| No                    | No                | No                |           |
+| Yes                   | No                | No                |0.75       |
+| No                    | 3D PCA            | No                |**0.93**   |
+| Yes                   | 3D PCA            | No                |0.86       |
+| Yes                   | No                | 10D PCA           |0.86       |
+| Yes                   | 3D PCA            | 10D PCA           |0.90       |
+
 
-| Left         | Centered         | Right                      |
-| :----------- | :--------------: | -------------------------: |
-| This is left | Text is centered | And this is right-aligned  |
-| More text    | Even more text   | And even more to the right |
+As we can see, the use of a ViT encoder does not necessarly improve the accuracy of the calssification.
+Indeed, the backbone used here was trained on RGB images, working on multispectral data seems to be to far a step here.
+Moreover, the different forest types have allready diffrente spectral responses.
 
+By desing, ViT are better than other ML methods to interpret the structure of an image and the relation between patches.
+Here, the information being mosty spectral leads to poorer performances after the use of a DL encoder, since the spectral information is drowned between unrelevant structural and tetural information interpreted by the DL backbone. 
+
+Here is the resulting map obtained when infering the RF model without pre-processing:
 
 ```{image} ./_static/examples/rf_no_red.png
 :alt: Random Forest inference without dimension reduction
@@ -27,6 +42,8 @@ Example table (temp)
 :align: center
 ```
 
+And with pre-processing:
+
 ```{image} ./_static/examples/rf_red.png
 :alt: Random Forest inference with dimension reduction
 :class: centered-image