Merge pull request microsoft#61 from microsoft/users/dan/conversion_f…

…ormat Users/dan/conversion format
vdedyukhin · Jun 22, 2019 · 021a392 · 021a392
2 parents 74054cd + aa0f523
commit 021a392
Show file tree

Hide file tree

Showing 2 changed files with 18 additions and 6 deletions.
diff --git a/api/batch_processing/postprocessing/convert_output_format.py b/api/batch_processing/postprocessing/convert_output_format.py
@@ -35,6 +35,16 @@ def convert_json_to_csv(input_path,output_path):
         detections = []
         for d in i['detections']:
             detection = d['bbox']
+
+            # Our .json format is xmin/ymin/w/h
+            #
+            # Our .csv format was ymin/xmin/ymax/xmax
+            xmin = detection[0]
+            ymin = detection[1]
+            xmax = detection[0] + detection[2]
+            ymax = detection[1] + detection[3]
+            detection = [ymin, xmin, ymax, xmax]
+
             detection.append(d['conf'])
             detection.append(int(d['category']))
             detections.append(detection)
@@ -79,12 +89,14 @@ def convert_csv_to_json(input_path,output_path):
 
         for iDetection,detection in enumerate(src_detections):
 
-            # Our .csv format was xmin/ymin/xmax/ymax
+            # Our .csv format was ymin/xmin/ymax/xmax
             #
             # Our .json format is xmin/ymin/w/h
-            bbox = detection[0:4]
-            bbox[2] = bbox[2] - bbox[0]
-            bbox[3] = bbox[3] - bbox[1]
+            ymin = detection[0]
+            xmin = detection[1]
+            ymax = detection[2]
+            xmax = detection[3]
+            bbox = [xmin, ymin, xmax-xmin, ymax-ymin]
             conf = detection[4]
             iClass = detection[5]
             out_detection = {}

diff --git a/visualization/visualization_utils.py b/visualization/visualization_utils.py
@@ -204,8 +204,8 @@ def render_detection_bounding_boxes_old(boxes_scores_classes, image,
     Args:
         boxes_and_scores:  outputs of generate_detections, in one of the following formats
         
-            [x_min, y_min, x_max, x_max, p]
-            [x_min, y_min, x_max, x_max, p, class]
+            [x_min, y_min, x_max, y_max, p]
+            [x_min, y_min, x_max, y_max, p, class]
             
         ...all in normalized coordinates, with the origin at the upper-left.