Add sweden data

data_prep/Radogoshi_Sweden.py

@@ -0,0 +1,94 @@
+import os
+import pandas as pd
+from deepforest.utilities import read_file
+import numpy as np
+import xmltodict
+
+
+def xml_to_annotations(xml_path, image_dir):
+    """
+    Load annotations from xml format (e.g. RectLabel editor) and convert
+    them into retinanet annotations format.
+    Args:
+        xml_path (str): Path to the annotations xml, formatted by RectLabel
+        image_dir: Directory to search for images
+    Returns:
+        Annotations (pandas dataframe): in the
+            format -> path-to-image.png,x1,y1,x2,y2,class_name
+    """
+    # parse
+    with open(xml_path) as fd:
+        doc = xmltodict.parse(fd.read())
+
+    # grab xml objects
+    try:
+        tile_xml = doc["annotation"]["object"]
+    except Exception as e:
+        raise Exception("error {} for path {} with doc annotation{}".format(
+            e, xml_path, doc["annotation"]))
+
+    xmin = []
+    xmax = []
+    ymin = []
+    ymax = []
+    label = []
+
+    if isinstance(tile_xml, list):
+        # Construct frame if multiple trees
+        for tree in tile_xml:
+            xmin.append(tree["bndbox"]["xmin"])
+            xmax.append(tree["bndbox"]["xmax"])
+            ymin.append(tree["bndbox"]["ymin"])
+            ymax.append(tree["bndbox"]["ymax"])
+    else:
+        xmin.append(tile_xml["bndbox"]["xmin"])
+        xmax.append(tile_xml["bndbox"]["xmax"])
+        ymin.append(tile_xml["bndbox"]["ymin"])
+        ymax.append(tile_xml["bndbox"]["ymax"])
+
+    rgb_name = os.path.basename(doc["annotation"]["filename"])
+
+    if os.path.exists("{}/{}".format(image_dir,rgb_name)):
+        # set dtypes, check for floats and round
+        xmin = [int(np.round(float(x))) for x in xmin]
+        xmax = [int(np.round(float(x)))for x in xmax]
+        ymin = [int(np.round(float(x))) for x in ymin]
+        ymax = [int(np.round(float(x))) for x in ymax]
+
+        annotations = pd.DataFrame({
+            "image_path": rgb_name,
+            "xmin": xmin,
+            "ymin": ymin,
+            "xmax": xmax,
+            "ymax": ymax,
+        })
+        return (annotations)
+    else:
+        raise IOError("{} doesn't exist".format(rgb_name))
+
+def recursive_search_and_process(root_dir):
+    xml_files = []
+    for subdir, _, files in os.walk(root_dir):
+        for file in files:
+            if file.endswith(".xml"):
+                xml_files.append(os.path.join(subdir, file))
+
+    dataframes = []
+    for xml_file in xml_files:
+        try:
+            df = xml_to_annotations(xml_file, root_dir + "/images")
+        except Exception as e:
+            print("Error processing file: {}. Error: {}".format(xml_file, e))
+            continue
+        df['image_path'] = os.path.join(root_dir + "/images", os.path.splitext(os.path.basename(xml_file))[0] + '.JPG')
+        df = read_file(df, root_dir + "/images")
+        dataframes.append(df)
+
+    combined_df = pd.concat(dataframes, ignore_index=True)
+    combined_df['label'] = 'Tree'
+    combined_df['source'] = 'Radogoshi et al. 2021'
+
+    combined_df.to_csv(os.path.join(root_dir, 'annotations.csv'), index=False)
+
+root_directory = "/orange/ewhite/DeepForest/Radogoshi_Sweden"
+recursive_search_and_process(root_directory)

data_prep/collect_tasks.py

@@ -12,6 +12,7 @@
     "/orange/ewhite/DeepForest/Ryoungseob_2023/train_datasets/images/train.csv",
     "/orange/ewhite/DeepForest/Velasquez_urban_trees/tree_canopies/nueva_carpeta/annotations.csv",
     '/orange/ewhite/DeepForest/individual_urban_tree_crown_detection/annotations.csv',
+    '/orange/ewhite/DeepForest/Radogoshi_Sweden/annotations.csv',
     '/orange/ewhite/DeepForest/ReForestTree/images/train.csv']
 
 TreePoints = [

docs/datasets.md

@@ -13,6 +13,14 @@ ISPRS Journal of Photogrammetry and Remote Sensing, Volume 206, 2023
 
 **Location:** Suwon, South Korea
 
+### Ragadoshi_Sweden
+
+![sample_image](public/Radogoshi_et_al._2021.png)
+
+**Link:** [https://lila.science/datasets/forest-damages-larch-casebearer/](https://lila.science/datasets/forest-damages-larch-casebearer/)
+
+**Location:** Sweden
+
 ### Velasquez-Camacho et al. 2023
 
 ![sample_image](public/Velasquez-Camacho_et_al._2023.png)
@@ -107,6 +115,8 @@ International Journal of Applied Earth Observation and Geoinformation, 130, 1038
 
 ### Safonova et al. 2021
 
+![sample_image](public/Safonova_et_al._2021.png)
+
 **Link:** [https://www.mdpi.com/1424-8220/21/5/1617](https://www.mdpi.com/1424-8220/21/5/1617)
 
 **Location:** Spain