Add latest model and error control

checkpoints/best_model.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:241e6ebaa818ecd1def1406bc3fa60702e49287e11bac30d4fcd8dd4b3c40b04
+oid sha256:5c4ffde314c0e590d8a4263c36acc4307c8035b343b27292f4de9fb871e6cfae
 size 21010087

utils.py

@@ -40,7 +40,10 @@ def remove_noise(points, eps=0.05, min_samples=10):
     """
     clustering = DBSCAN(eps=eps, min_samples=min_samples).fit(points)
     labels = clustering.labels_
-    largest_cluster = labels == np.argmax(np.bincount(labels[labels >= 0]))
+    try:
+        largest_cluster = labels == np.argmax(np.bincount(labels[labels >= 0]))
+    except Exception:
+        raise RuntimeError("Error in DBH calculation")
     return points[largest_cluster]
 
 def calculate_dbh(points, dbh_height=1.3, height_buffer=0.1, eps=0.05, min_samples=10):
@@ -95,9 +98,8 @@ def calc_canopy_volume(points, threshold, height, z_min):
     Returns:
         canopy_volume, canopy_points
     '''
-    canopy_points = points[points[:, 2] > z_min + 0.2 * height]
-    z_threshold = np.percentile(canopy_points[:, 2], threshold)
-    canopy_points = canopy_points[canopy_points[:, 2] >= z_threshold]
+    z_threshold = z_min + (threshold / 100) * height
+    canopy_points = points[points[:, 2] >= z_threshold]
     clustering = DBSCAN(eps=1.0, min_samples=10).fit(canopy_points[:, :3])
     labels = clustering.labels_
     canopy_points = canopy_points[labels != -1]