bug: ground truth generation and proposal alignment

src/deep_neurographs/densegraph.py

@@ -26,14 +26,14 @@ class DenseGraph:
 
     """
 
-    def __init__(self, swc_dir):
+    def __init__(self, swc_paths):
         """
         Constructs a DenseGraph object from a directory of swc files.
 
         Parameters
         ----------
-        swc_dir : path
-            Path to directory of swc files which are used to construct a hash
+        swc_paths : list[str]
+            List of paths to swc files which are used to construct a hash
             table in which the entries are filename-graph pairs.
 
         Returns
@@ -43,18 +43,18 @@ def __init__(self, swc_dir):
         """
         self.xyz_to_node = dict()
         self.xyz_to_swc = dict()
-        self.init_graphs(swc_dir)
+        self.init_graphs(swc_paths)
         self.init_kdtree()
 
-    def init_graphs(self, swc_dir):
+    def init_graphs(self, swc_paths):
         """
-        Initializes graphs by reading swc files in "swc_dir". Graphs are
+        Initializes graphs by reading swc files in "swc_paths". Graphs are
         stored in a hash table where the entries are filename-graph pairs.
 
         Parameters
         ----------
-        swc_dir : path
-            Path to directory of swc files which are used to construct a hash
+        swc_paths : list[str]
+            List of paths to swc files which are used to construct a hash
             table in which the entries are filename-graph pairs.
 
         Returns
@@ -63,16 +63,15 @@ def init_graphs(self, swc_dir):
 
         """
         self.graphs = dict()
-        for f in utils.listdir(swc_dir, ext=".swc"):
+        for path in swc_paths:
             # Construct Graph
-            path = os.path.join(swc_dir, f)
             swc_dict = swc_utils.parse_local_swc(path)
             graph, xyz_to_node = swc_utils.to_graph(swc_dict, set_attrs=True)
 
             # Store
-            xyz_to_id = dict(zip_broadcast(swc_dict["xyz"], f))
-            self.graphs[f] = graph
-            self.xyz_to_node[f] = xyz_to_node
+            xyz_to_id = dict(zip_broadcast(swc_dict["xyz"], path))
+            self.graphs[path] = graph
+            self.xyz_to_node[path] = xyz_to_node
             self.xyz_to_swc.update(xyz_to_id)
 
     def init_kdtree(self):

src/deep_neurographs/intake.py

@@ -65,6 +65,7 @@ def build_neurograph_from_local(
         swc_dicts,
         bbox=bbox,
         img_path=img_path,
+        swc_paths=paths,
         prune=prune,
         prune_depth=prune_depth,
         smooth=smooth,
@@ -245,36 +246,11 @@ def list_gcs_filenames(bucket, cloud_path, extension):
 
 
 # -- Build neurograph ---
-def build_neurograph_old(
-    swc_dicts,
-    bbox=None,
-    img_path=None,
-    prune=PRUNE,
-    prune_depth=PRUNE_DEPTH,
-    smooth=SMOOTH,
-):
-    # Extract irreducibles
-    t0 = time()
-    irreducibles = dict()
-    for key in swc_dicts.keys():
-        irreducibles[key] = gutils.get_irreducibles(
-            swc_dicts[key], prune=prune, depth=prune_depth, smooth=smooth
-        )
-    print(f"   --> get_irreducibles(): {time() - t0} seconds")
-
-    # Build neurograph
-    t0 = time()
-    neurograph = NeuroGraph(bbox=bbox, img_path=img_path)
-    for key in swc_dicts.keys():
-        neurograph.add_immutables(swc_dicts[key], irreducibles[key])
-    print(f"   --> add_irreducibles(): {time() - t0} seconds")
-    return neurograph
-
-
 def build_neurograph(
     swc_dicts,
     bbox=None,
     img_path=None,
+    swc_paths=None,
     prune=PRUNE,
     prune_depth=PRUNE_DEPTH,
     smooth=SMOOTH,
@@ -310,7 +286,7 @@ def build_neurograph(
 
     # Build neurograph
     t0 = time()
-    neurograph = NeuroGraph(bbox=bbox, img_path=img_path)
+    neurograph = NeuroGraph(bbox=bbox, img_path=img_path, swc_paths=swc_paths)
     for key in swc_dicts.keys():
         neurograph.add_immutables(irreducibles[key], swc_dicts[key], key)
     print(f"   --> add_irreducibles(): {time() - t0} seconds")

src/deep_neurographs/neurograph.py

@@ -33,13 +33,13 @@ class NeuroGraph(nx.Graph):
     """
 
     def __init__(
-        self, bbox=None, swc_dir=None, img_path=None, label_mask=None
+        self, bbox=None, swc_paths=None, img_path=None, label_mask=None
     ):
         super(NeuroGraph, self).__init__()
         # Initialize paths
         self.img_path = img_path
         self.label_mask = label_mask
-        self.swc_paths = swc_dir
+        self.swc_paths = swc_paths
 
         # Initialize node and edge sets
         self.leafs = set()
@@ -204,7 +204,7 @@ def __get_proposals(
         best_dist = dict()
         query_swc_id = self.nodes[query_id]["swc_id"]
         for xyz in self._query_kdtree(query_xyz, search_radius):
-            if not self.is_contained(xyz):
+            if not self.is_contained(xyz, buffer=36):
                 continue
             xyz = tuple(xyz)
             edge = self.xyz_to_edge[xyz]
@@ -336,8 +336,8 @@ def orient_edge(self, edge, i):
     # --- Ground Truth Generation ---
     def init_targets(self, target_neurograph):
         # Initializations
-        msg = "Error: Provide swc_dir/swc_paths to initialize target edges!"
-        assert target_neurograph.swc_path, msg
+        msg = "Provide swc_dir/swc_paths to initialize target edges!"
+        assert target_neurograph.swc_paths, msg
         target_neurograph.init_densegraph()
         target_neurograph.init_kdtree()
         self.target_edges = set()
@@ -534,11 +534,11 @@ def get_projection(self, xyz):
     def is_nb(self, i, j):
         return True if i in self.neighbors(j) else False
 
-    def is_contained(self, node_or_xyz):
+    def is_contained(self, node_or_xyz, buffer=0):
         if self.bbox:
             if type(node_or_xyz) == int:
                 node_or_xyz = deepcopy(self.nodes[node_or_xyz]["xyz"])
-            return utils.is_contained(self.bbox, node_or_xyz)
+            return utils.is_contained(self.bbox, node_or_xyz, buffer=buffer)
         else:
             return True
 

src/deep_neurographs/swc_utils.py

@@ -21,15 +21,16 @@
 def process_local_paths(paths, min_size, bbox=None):
     swc_dicts = dict()
     for path in paths:
-        swc_dicts.update(parse_local_swc(path, bbox=bbox, min_size=min_size))
+        swc_dict = parse_local_swc(path, bbox=bbox, min_size=min_size)
+        if len(swc_dict):
+            swc_id = utils.get_swc_id(path)
+            swc_dicts[swc_id] = swc_dict
     return swc_dicts
 
 
-def parse_local_swc(path, bbox=None, min_size=25):
-    swc_id = utils.get_swc_id(path)
-    swc_contents = read_from_local(path)
-    swc_dict = parse(swc_contents, bbox=bbox)
-    return {swc_id: swc_dict} if len(swc_dict["id"]) > min_size else dict()
+def parse_local_swc(path, bbox=None, min_size=0):
+    contents = read_from_local(path)
+    return parse(contents, bbox=bbox) if len(contents) > min_size else []
 
 
 def parse_gcs_zip(zip_file, path, min_size=0):
@@ -80,7 +81,7 @@ def parse(swc_contents, bbox=None):
         swc_dict["id"][i] -= min_id
         swc_dict["pid"][i] -= min_id
 
-    return swc_dict
+    return swc_dict if len(swc_dict["id"]) > 1 else []
 
 
 def read_from_local(path):

src/deep_neurographs/utils.py

@@ -411,10 +411,10 @@ def get_avg_std(data, weights=None):
     return avg, math.sqrt(var)
 
 
-def is_contained(bbox, xyz):
+def is_contained(bbox, xyz, buffer=0):
     xyz = apply_anisotropy(xyz - bbox["min"])
     shape = bbox["max"] - bbox["min"]
-    if any(xyz < 0) or any(xyz >= shape):
+    if any(xyz < buffer) or any(xyz >= shape - buffer):
         return False
     else:
         return True