Use npy in labelme2voc.py for -1 value in labels

PNG does not keep negative values.

examples/instance_segmentation/labelme2voc.py

@@ -61,11 +61,11 @@ def main():
             out_img_file = osp.join(
                 args.out_dir, 'JPEGImages', base + '.jpg')
             out_cls_file = osp.join(
-                args.out_dir, 'SegmentationClass', base + '.png')
+                args.out_dir, 'SegmentationClass', base + '.npy')
             out_clsv_file = osp.join(
                 args.out_dir, 'SegmentationClassVisualization', base + '.jpg')
             out_ins_file = osp.join(
-                args.out_dir, 'SegmentationObject', base + '.png')
+                args.out_dir, 'SegmentationObject', base + '.npy')
             out_insv_file = osp.join(
                 args.out_dir, 'SegmentationObjectVisualization', base + '.jpg')
 
@@ -81,15 +81,16 @@ def main():
                 label_name_to_value=class_name_to_id,
                 type='instance',
             )
+            ins[cls == -1] = 0  # ignore it.
 
-            PIL.Image.fromarray(cls).save(out_cls_file)
+            np.save(out_cls_file, cls)
             label_names = ['%d: %s' % (cls_id, cls_name)
                            for cls_id, cls_name in enumerate(class_names)]
             clsv = labelme.utils.draw_label(
                 cls, img, label_names, colormap=colormap)
             PIL.Image.fromarray(clsv).save(out_clsv_file)
 
-            PIL.Image.fromarray(ins).save(out_ins_file)
+            np.save(out_ins_file, ins)
             instance_ids = np.unique(ins)
             instance_names = [str(i) for i in range(max(instance_ids) + 1)]
             insv = labelme.utils.draw_label(

examples/semantic_segmentation/labelme2voc.py

@@ -59,7 +59,7 @@ def main():
             out_img_file = osp.join(
                 args.out_dir, 'JPEGImages', base + '.jpg')
             out_lbl_file = osp.join(
-                args.out_dir, 'SegmentationClass', base + '.png')
+                args.out_dir, 'SegmentationClass', base + '.npy')
             out_viz_file = osp.join(
                 args.out_dir, 'SegmentationClassVisualization', base + '.jpg')
 
@@ -75,11 +75,10 @@ def main():
                 label_name_to_value=class_name_to_id,
             )
 
-            lbl_pil = PIL.Image.fromarray(lbl)
             # Only works with uint8 label
             # lbl_pil = PIL.Image.fromarray(lbl, mode='P')
             # lbl_pil.putpalette((colormap * 255).flatten())
-            lbl_pil.save(out_lbl_file)
+            np.save(out_lbl_file, lbl)
 
             label_names = ['%d: %s' % (cls_id, cls_name)
                            for cls_id, cls_name in enumerate(class_names)]