renaming class identifiers functionality in set_seg_options

kraken/contrib/new_classes.json

@@ -0,0 +1 @@
+{"baselines": {"defaultLine": 2}, "regions": {"foo": 3}}

kraken/contrib/set_seg_options.py

@@ -12,41 +12,66 @@
 @click.option('--topline/--baseline', default=False, help='Sets model line type to baseline or topline')
 @click.option('--pad', show_default=True, type=(int, int), default=(0, 0),
               help='Padding (left/right, top/bottom) around the page image')
+@click.option('--output-identifiers', type=click.Path(exists=True), help='Path '
+              'to a json file containing a dict updating the string identifiers '
+              'of line/region classes.')
 @click.argument('model', nargs=1, type=click.Path(exists=True))
-def cli(bounding_region, topline, pad, model):
+def cli(bounding_region, topline, pad, output_identifiers, model):
     """
     A script setting the metadata of segmentation models.
     """
+    import json
     from kraken.lib import vgsl
 
     net = vgsl.TorchVGSLModel.load_model(model)
     if net.model_type != 'segmentation':
         print('Model is not a segmentation model.')
         return
 
-    print('detectable line and region types:')
+    print('detectable line types:')
     for k, v in net.user_metadata['class_mapping']['baselines'].items():
         print(f'  {k}\t{v}')
-    print('Training region types:')
+    print('detectable region types:')
     for k, v in net.user_metadata['class_mapping']['regions'].items():
         print(f'  {k}\t{v}')
 
+    if output_identifiers:
+        with open(output_identifiers, 'r') as fp:
+            new_cls_map = json.load(fp)
+        print('-> Updating class maps')
+        if 'baselines' in new_cls_map:
+            print('new baseline identifiers:')
+            old_cls = {v: k for k,v in net.user_metadata['class_mapping']['baselines'].items()}
+            new_cls = {v: k for k,v in new_cls_map['baselines'].items()}
+            old_cls.update(new_cls)
+            net.user_metadata['class_mapping']['baselines'] = {v: k for k, v in old_cls.items()}
+            for k, v in net.user_metadata['class_mapping']['baselines'].items():
+                print(f'  {k}\t{v}')
+        if 'regions' in new_cls_map:
+            print('new region identifiers:')
+            old_cls = {v: k for k,v in net.user_metadata['class_mapping']['regions'].items()}
+            new_cls = {v: k for k,v in new_cls_map['regions'].items()}
+            old_cls.update(new_cls)
+            net.user_metadata['class_mapping']['regions'] = {v: k for k, v in old_cls.items()}
+            for k, v in net.user_metadata['class_mapping']['regions'].items():
+                print(f'  {k}\t{v}')
+
     print(f'existing bounding regions: {net.user_metadata["bounding_regions"]}')
 
     if bounding_region:
         br = set(net.user_metadata["bounding_regions"])
         br_new = set(bounding_region)
 
-        print(f'removing: {br.difference(br_new)}')
-        print(f'adding: {br_new.difference(br)}')
+        print(f'-> removing: {br.difference(br_new)}')
+        print(f'-> adding: {br_new.difference(br)}')
         net.user_metadata["bounding_regions"] = bounding_region
 
     print(f'Model is {"topline" if "topline" in net.user_metadata and net.user_metadata["topline"] else "baseline"}')
-    print(f'Setting to  {"topline" if topline else "baseline"}')
+    print(f'-> Setting to  {"topline" if topline else "baseline"}')
     net.user_metadata['topline'] = topline
 
     print(f"Model has padding {net.user_metadata['hyper_params']['padding'] if 'padding' in net.user_metadata['hyper_params'] else (0, 0)}")
-    print(f'Setting to {pad}')
+    print(f'-> Setting to {pad}')
     net.user_metadata['hyper_params']['padding'] = pad
 
     shutil.copy(model, f'{model}.bak')