Debug metrics

helpers/metrics.py

@@ -150,15 +150,7 @@ def get_metrics(tp_gdf, fp_gdf, fn_gdf, mismatch_gdf, id_classes=0, method='macr
     pw_k = by_class_dict.copy()
     rw_k = by_class_dict.copy()
 
-    by_class_dict = {key: None for key in id_classes}
-    tp_k = by_class_dict.copy()
-    fp_k = by_class_dict.copy()
-    fn_k = by_class_dict.copy()
-    p_k = by_class_dict.copy()
-    r_k = by_class_dict.copy()
-    count_k = by_class_dict.copy()
-    pw_k = by_class_dict.copy()
-    rw_k = by_class_dict.copy()
+    total_labels = len(tp_gdf) + len(fn_gdf) + len(mismatch_gdf)
 
     for id_cl in id_classes:
 
@@ -176,25 +168,27 @@ def get_metrics(tp_gdf, fp_gdf, fn_gdf, mismatch_gdf, id_classes=0, method='macr
         fn_count = pure_fn_count + mismatched_fn_count
         tp_count = 0 if tp_gdf.empty else len(tp_gdf[tp_gdf.det_class==id_cl])
 
-        tp_k[id_cl] = tp_count
         fp_k[id_cl] = fp_count
         fn_k[id_cl] = fn_count
-
-        p_k[id_cl] = 0 if tp_count == 0 else tp_count / (tp_count + fp_count)
-        r_k[id_cl] = 0 if tp_count == 0 else tp_count / (tp_count + fn_count)
-        count_k[id_cl] = 0 if tp_count == 0 else tp_count + fn_count 
+        tp_k[id_cl] = tp_count
+
+        count_k[id_cl] = tp_count + fn_count
+        if tp_count > 0:
+            p_k[id_cl] = tp_count / (tp_count + fp_count)
+            r_k[id_cl] = tp_count / (tp_count + fn_count)
+
+        if (method == 'macro-weighted-average') & (total_labels > 0):
+            pw_k[id_cl] = (count_k[id_cl] / total_labels) * p_k[id_cl]
+            rw_k[id_cl] = (count_k[id_cl] / total_labels) * r_k[id_cl] 
 
     if method == 'macro-average':   
         precision = sum(p_k.values()) / len(id_classes)
         recall = sum(r_k.values()) / len(id_classes)
-    elif method == 'macro-weighted-average': 
-        for id_cl in id_classes:
-            pw_k[id_cl] = 0 if sum(count_k.values()) == 0 else (count_k[id_cl] / sum(count_k.values())) * p_k[id_cl]
-            rw_k[id_cl] = 0 if sum(count_k.values()) == 0 else (count_k[id_cl] / sum(count_k.values())) * r_k[id_cl] 
+    elif method == 'macro-weighted-average':  
         precision = sum(pw_k.values()) / len(id_classes)
         recall = sum(rw_k.values()) / len(id_classes)
     elif method == 'micro-average':  
-        if sum(tp_k.values()) == 0 and sum(fp_k.values()) == 0:
+        if sum(tp_k.values()) == 0:
             precision = 0
             recall = 0
         else:
@@ -224,4 +218,9 @@ def intersection_over_union(polygon1_shape, polygon2_shape):
     polygon_intersection = polygon1_shape.intersection(polygon2_shape).area
     polygon_union = polygon1_shape.area + polygon2_shape.area - polygon_intersection
 
-    return polygon_intersection / polygon_union
+    if polygon_union != 0:
+        iou = polygon_intersection / polygon_union
+    else:
+        iou = 0
+
+    return iou

helpers/misc.py

@@ -16,6 +16,15 @@
 from shapely.validation import make_valid
 from rasterio.transform import from_bounds
 
+try:
+    try:
+        from helpers.metrics import intersection_over_union
+    except ModuleNotFoundError:
+        from metrics import intersection_over_union
+except Exception as e:
+    logger.error(f"Could not import some dependencies. Exception: {e}")
+    sys.exit(1)
+
 
 class BadFileExtensionException(Exception):
     "Raised when the file extension is different from the expected one"
@@ -144,29 +153,6 @@ def clip_row(row, fact=fact):
     return clipped_labels_gdf
 
 
-def intersection_over_union(polygon1_shape, polygon2_shape):
-    """Determine the intersection area over union area (IoU) of two polygons
-
-    Args:
-        polygon1_shape (geometry): first polygon
-        polygon2_shape (geometry): second polygon
-
-    Returns:
-        int: Unrounded ratio between the intersection and union area
-    """
-
-    # Calculate intersection and union, and the IoU
-    polygon_intersection = polygon1_shape.intersection(polygon2_shape).area
-    polygon_union = polygon1_shape.area + polygon2_shape.area - polygon_intersection
-
-    if polygon_union != 0:
-        iou = polygon_intersection / polygon_union
-    else:
-        iou = 0
-
-    return iou
-
-
 def format_logger(logger):
 
     logger.remove()

scripts/make_detections.py

@@ -62,6 +62,7 @@ def main(cfg_file_path):
     DETECTRON2_CFG_FILE = cfg['detectron2_config_file']
 
     WORKING_DIR = cfg['working_directory']
+    OUTPUT_DIR = cfg['output_folder'] if 'output_folder' in cfg.keys() else '.'
     SAMPLE_TAGGED_IMG_SUBDIR = cfg['sample_tagged_img_subfolder']
     LOG_SUBDIR = cfg['log_subfolder']
 
@@ -74,9 +75,8 @@ def main(cfg_file_path):
 
     os.chdir(WORKING_DIR)
     # let's make the output directories in case they don't exist
-    for DIR in [SAMPLE_TAGGED_IMG_SUBDIR, LOG_SUBDIR]:
-        if not os.path.exists(DIR):
-            os.makedirs(DIR)
+    for directory in [OUTPUT_DIR, SAMPLE_TAGGED_IMG_SUBDIR, LOG_SUBDIR]:
+        os.makedirs(directory, exist_ok=True)
 
     written_files = []
 
@@ -121,7 +121,7 @@ def main(cfg_file_path):
 
         logger.info(f"Making detections over the entire {dataset} dataset...")
 
-        detections_filename = f'{dataset}_detections_at_{threshold_str}_threshold.gpkg'
+        detections_filename = os.path.join(OUTPUT_DIR, f'{dataset}_detections_at_{threshold_str}_threshold.gpkg')
 
         for d in tqdm(DatasetCatalog.get(dataset)):
 
@@ -181,13 +181,14 @@ def main(cfg_file_path):
             im = cv2.imread(d["file_name"])
             outputs = predictor(im)
             v = Visualizer(im[:, :, ::-1], # [:, :, ::-1] is for RGB -> BGR conversion, cf. https://stackoverflow.com/questions/14556545/why-opencv-using-bgr-colour-space-instead-of-rgb
-                           metadata=MetadataCatalog.get(dataset), 
-                           scale=1.0, 
-                           instance_mode=ColorMode.IMAGE_BW # remove the colors of unsegmented pixels
+                metadata=MetadataCatalog.get(dataset), 
+                scale=1.0, 
+                instance_mode=ColorMode.IMAGE_BW # remove the colors of unsegmented pixels
             )   
             v = v.draw_instance_predictions(outputs["instances"].to("cpu"))
-            cv2.imwrite(os.path.join(SAMPLE_TAGGED_IMG_SUBDIR, output_filename), v.get_image()[:, :, ::-1])
-            written_files.append(os.path.join(WORKING_DIR, SAMPLE_TAGGED_IMG_SUBDIR, output_filename))
+            filepath = os.path.join(SAMPLE_TAGGED_IMG_SUBDIR, output_filename)
+            cv2.imwrite(filepath, v.get_image()[:, :, ::-1])
+            written_files.append(os.path.join(WORKING_DIR, filepath))
         logger.success(DONE_MSG)