Update README.md

README.md

@@ -45,7 +45,7 @@ model-index:
   <h1>FRACTAL-LidarHD_7cl_randlanet</h1> 
   <p>The general characteristics of this specific model <strong>FRACTAL-LidarHD_7cl_randlanet</strong> are :</p>
   <ul style="list-style-type:disc;">
-    <li>Trained with the FRACTAL dataset</li>
+    <li>Trained with the FRACTAL dataset for the semantic segmentation of Lidar HD point clouds</li>
     <li>Aerial lidar point clouds, colorized with rgb + near-infrared, with high point density (~40 pts/m²)</li>
     <li>RandLa-Net architecture as implemented in the Myria3D library</li>
     <li>7 class nomenclature : [other, ground, vegetation, building, water, bridge, permanent structure]</li>
@@ -157,12 +157,39 @@ The model was obtained for num_epoch=21 with corresponding val_loss=0.112.
 
 #### Testing Data
 
+The model was evaluated on the 10,000 data patches of the test set of the FRACTAL dataset, 
+that are independant from train and val patches, and sampled from distinct areas in the five spatial domains of the dataset.
+The diversity of landscapes and scenes of the test set should closely match the one of the train and val sets.
+
 #### Metrics
 
+The **FRACTAL-LidarHD_7cl_randlanet** model achieves a performance of mIoU=77.2% and OA=XXX%.
+
+The following table fives the class-wise metrics:
+
+TODO: add IoU, and other metrics if available.
+
+
+The following illustration gives the resulting confusion matrix :
+* Left : normalised acording to rows: rows sum at 100% and the **recall** is on the diagonal of the matrix 
+* Right : normalised acording to columns: columns sum at 100% and the **precision** is on the diagonal of the matrix
+
+
+<div style="position: relative; text-align: center;">
+    <p style="margin: 0;">Normalized Confusion Matrices. (a) Recall, (b) Precision)</p>
+    <img src="FRACTAL-LidarHD_7cl_randlanet-recall_confusion_matrix.excalidraw.png" alt="Confusion matrices" style="width: 70%; display: block; margin: 0 auto;"/>
+</div>
+
 
 ### Results
 
-Samples of results
+Samples of results. 
+From test patches with at least 10k points (i.e. at least 4 pts/m²), we sample without cherry-picking, with the following criteria: 
+* (1) WATER & BRIDGE
+* (2) BUILD_GREENHOUSE
+* (3) OTHER_PARKING
+* (4) HIGHSLOPE1
+* (5) URBAN
 
 
 ---