Fix case of missing label in Forest Damage dataset (microsoft#499)

* fix case of missing label

* mypy fix

tests/data/forestdamage/Data_Set_Larch_Casebearer/Bebehojd_20190527/Annotations/B01_0004.xml

@@ -1 +1 @@
-<annotation><filename>B01_0004.xml</filename><size><width>32</width><height>32</height><depth>3</depth></size><object><damage>other</damage><bndbox><xmin>8</xmin><ymin>8</ymin><xmax>24</xmax><ymax>24</ymax></bndbox></object></annotation>
+<annotation><filename>B01_0004.xml</filename><size><width>32</width><height>32</height><depth>3</depth></size><object><damage>other</damage><bndbox><xmin>8</xmin><ymin>8</ymin><xmax>24</xmax><ymax>24</ymax></bndbox></object><object><bndbox><xmin>8</xmin><ymin>8</ymin><xmax>24</xmax><ymax>24</ymax></bndbox></object></annotation>

tests/data/forestdamage/Data_Set_Larch_Casebearer/Bebehojd_20190527/Annotations/B01_0005.xml

@@ -1 +1 @@
-<annotation><filename>B01_0005.xml</filename><size><width>32</width><height>32</height><depth>3</depth></size><object><damage>other</damage><bndbox><xmin>8</xmin><ymin>8</ymin><xmax>24</xmax><ymax>24</ymax></bndbox></object></annotation>
+<annotation><filename>B01_0005.xml</filename><size><width>32</width><height>32</height><depth>3</depth></size><object><damage>other</damage><bndbox><xmin>8</xmin><ymin>8</ymin><xmax>24</xmax><ymax>24</ymax></bndbox></object><object><bndbox><xmin>8</xmin><ymin>8</ymin><xmax>24</xmax><ymax>24</ymax></bndbox></object></annotation>

tests/data/forestdamage/data.py

@@ -24,6 +24,7 @@
         "Bebehojd_20190527/Annotations/B01_0004.xml",
         "Bebehojd_20190527/Annotations/B01_0005.xml",
     ],
+    "labels": [True, False],
 }
 
 
@@ -38,15 +39,17 @@ def create_annotation(path: str) -> None:
     ET.SubElement(size, "height").text = str(SIZE)
     ET.SubElement(size, "depth").text = str(3)
 
-    annotation = ET.SubElement(root, "object")
+    for label in PATHS["labels"]:
+        annotation = ET.SubElement(root, "object")
 
-    ET.SubElement(annotation, "damage").text = "other"
+        if label:
+            ET.SubElement(annotation, "damage").text = "other"
 
-    bbox = ET.SubElement(annotation, "bndbox")
-    ET.SubElement(bbox, "xmin").text = str(0 + int(SIZE / 4))
-    ET.SubElement(bbox, "ymin").text = str(0 + int(SIZE / 4))
-    ET.SubElement(bbox, "xmax").text = str(SIZE - int(SIZE / 4))
-    ET.SubElement(bbox, "ymax").text = str(SIZE - int(SIZE / 4))
+        bbox = ET.SubElement(annotation, "bndbox")
+        ET.SubElement(bbox, "xmin").text = str(0 + int(SIZE / 4))
+        ET.SubElement(bbox, "ymin").text = str(0 + int(SIZE / 4))
+        ET.SubElement(bbox, "xmax").text = str(SIZE - int(SIZE / 4))
+        ET.SubElement(bbox, "ymax").text = str(SIZE - int(SIZE / 4))
 
     tree = ET.ElementTree(root)
     tree.write(path)

tests/datasets/test_forestdamage.py

@@ -27,7 +27,7 @@ def dataset(self, monkeypatch: MonkeyPatch, tmp_path: Path) -> ForestDamage:
 
         url = os.path.join(data_dir, "Data_Set_Larch_Casebearer.zip")
 
-        md5 = "a6adc19879c1021cc1ba8d424e19c9e0"
+        md5 = "52d82ac38899e6e6bb40aacda643ee15"
 
         monkeypatch.setattr(ForestDamage, "url", url)
         monkeypatch.setattr(ForestDamage, "md5", md5)

torchgeo/datasets/forestdamage.py

@@ -40,7 +40,12 @@ def parse_pascal_voc(path: str) -> Dict[str, Any]:
             int(bndbox.find("xmax").text),  # type: ignore[union-attr, arg-type]
             int(bndbox.find("ymax").text),  # type: ignore[union-attr, arg-type]
         ]
-        label = obj.find("damage").text  # type: ignore[union-attr]
+
+        label_var = obj.find("damage")
+        if label_var is not None:
+            label = label_var.text
+        else:
+            label = "other"
         bboxes.append(bbox)
         labels.append(label)
     return dict(filename=filename, bboxes=bboxes, labels=labels)