Add optional license and info

pyodi/apps/coco/coco_split.py

@@ -133,15 +133,15 @@ def property_split(
     train_split = {
         "images": train_images,
         "annotations": train_annotations,
-        "info": data["info"],
-        "licenses": data["licenses"],
+        "info": data.get("info", {}),
+        "licenses": data.get("licenses", []),
         "categories": data["categories"],
     }
     val_split = {
         "images": val_images,
         "annotations": val_annotations,
-        "info": data["info"],
-        "licenses": data["licenses"],
+        "info": data.get("info", {}),
+        "licenses": data.get("licenses", []),
         "categories": data["categories"],
     }
 
@@ -202,16 +202,16 @@ def random_split(
     train_split = {
         "images": train_images,
         "annotations": train_annotations,
-        "info": data["info"],
-        "licenses": data["licenses"],
+        "info": data.get("info", {}),
+        "licenses": data.get("licenses", []),
         "categories": data["categories"],
     }
 
     val_split = {
         "images": val_images,
         "annotations": val_annotations,
-        "info": data["info"],
-        "licenses": data["licenses"],
+        "info": data.get("info", {}),
+        "licenses": data.get("licenses", []),
         "categories": data["categories"],
     }
 

tests/apps/test_coco_split.py

@@ -1,6 +1,8 @@
 import json
 from pathlib import Path
 
+import pytest
+
 from pyodi.apps.coco.coco_split import property_split, random_split
 
 
@@ -102,3 +104,41 @@ def test_property_coco_split(tmpdir):
     assert len(val_data["images"]) == 4
     assert len(train_data["annotations"]) == 1
     assert len(val_data["annotations"]) == 8
+
+
+@pytest.mark.parametrize("split_type", ["random", "property"])
+def test_split_without_info_and_licenses(tmpdir, split_type):
+    tmpdir = Path(tmpdir)
+
+    coco_data = get_coco_data()
+    coco_data.pop("licenses")
+    coco_data.pop("info")
+
+    assert "licenses" not in coco_data
+    assert "info" not in coco_data
+
+    json.dump(coco_data, open(tmpdir / "coco.json", "w"))
+
+    if split_type == "random":
+        train_path, val_path = random_split(
+            annotations_file=str(tmpdir / "coco.json"),
+            output_filename=str(tmpdir / "random_coco_split"),
+            val_percentage=0.25,
+            seed=49,
+        )
+    else:
+        config = dict(
+            val={"file_name": {"frame 0": "vidA-0.jpg", "frame 1": "vidA-1.jpg"}}
+        )
+        json.dump(config, open(tmpdir / "split_config.json", "w"))
+
+        train_path, val_path = property_split(
+            annotations_file=str(tmpdir / "coco.json"),
+            output_filename=str(tmpdir / "property_coco_split"),
+            split_config_file=str(tmpdir / "split_config.json"),
+        )
+
+    for path in [train_path, val_path]:
+        data = json.load(open(path))
+        assert data["licenses"] == []
+        assert data["info"] == {}