Wagner added

data_prep/Firoze2023.py

@@ -0,0 +1,38 @@
+import os
+import json
+import pandas as pd
+
+def extract_bounding_boxes_from_labelme(folder_path):
+    data = []
+
+    for filename in os.listdir(folder_path):
+        if filename.endswith('.json'):
+            json_path = os.path.join(folder_path, filename)
+            with open(json_path, 'r') as file:
+                labelme_data = json.load(file)
+                image_path = os.path.join(folder_path, labelme_data['imagePath'])
+
+                polygons = []
+                for shape in labelme_data['shapes']:
+                    points = shape['points']
+                    # Add the first to the end to close the polygon
+                    points.append(points[0])
+                    wkt_polygon = 'POLYGON(({}))'.format(', '.join(['{} {}'.format(p[0], p[1]) for p in points]))
+                    polygons.append([image_path, wkt_polygon])
+
+                # Create dataframe with image path
+                df = pd.DataFrame(polygons, columns=['image_path', 'geometry'])
+                data.append(df)
+
+    annotations = pd.concat(data)
+
+    return annotations
+
+# Example usage
+folder_path = '/orange/ewhite/DeepForest/Firoze2023/annotated_forest_dataset/annotated_real_forest'
+df = extract_bounding_boxes_from_labelme(folder_path)
+df["source"] = "Firoze et al. 2023"
+df["label"] = "Tree"
+print("There are {} annotations in {} images".format(df.shape[0], len(df.image_path.unique())))
+df.to_csv("/orange/ewhite/DeepForest/Firoze2023/annotations.csv")
+

data_prep/Wagner.py

@@ -0,0 +1,140 @@
+import os
+import geopandas as gpd
+from shapely.geometry import box
+from deepforest.utilities import read_file
+import pandas as pd
+import rasterio as rio
+from rasterio.warp import calculate_default_transform, reproject, Resampling
+
+
+# Define paths
+geopackage_dir = "/orange/ewhite/DeepForest/Wagner_Australia/Crowns (manual)"
+cropped_plots_dir = "/orange/ewhite/DeepForest/Wagner_Australia/Cropped plots"
+output_csv = "/orange/ewhite/DeepForest/Wagner_Australia/annotations.csv"
+
+# Padding in meters
+padding = 10
+
+# Initialize an empty list to store annotations
+annotations = []
+
+# Iterate over each geopackage in the directory
+for filename in os.listdir(geopackage_dir):
+    if filename.endswith(".gpkg"):
+        basename = os.path.splitext(filename)[0]
+        geopackage_path = os.path.join(geopackage_dir, filename)
+
+        basename = basename.replace(".", "")
+
+        # Look for the corresponding image in the cropped plots folder
+        image_path = os.path.join(cropped_plots_dir, basename + ".tif")
+        if not os.path.exists(image_path):
+            print(f"Image for {basename} not found in cropped plots folder.")
+            continue
+
+        # Read the geopackage
+        gdf = gpd.read_file(geopackage_path)
+
+        # Get the total bounds of the polygons and pad by a few meters
+        minx, miny, maxx, maxy = gdf.total_bounds
+        minx -= padding
+        miny -= padding
+        maxx += padding
+        maxy += padding
+
+        # Read the image using rasterio
+        with rio.open(image_path) as src:
+            # Read the first three bands
+            bands = src.read([1, 2, 3])
+
+            # Define the output path for the three-band image
+            three_band_path = os.path.join("/orange/ewhite/DeepForest/Wagner_Australia/three_band", basename + "_three_band.tif")
+
+            # Save the three-band image
+
+            # project into utm S55
+
+            # Save the three-band image
+            with rio.open(
+                three_band_path,
+                'w',
+                driver='GTiff',
+                height=src.height,
+                width=src.width,
+                count=3,
+                dtype=bands.dtype,
+                crs=src.crs,
+                transform=src.transform,
+            ) as dst:
+                dst.write(bands[0], 1)
+                dst.write(bands[1], 2)
+                dst.write(bands[2], 3)
+
+            # Reproject the three-band image to UTM S55
+            with rio.open(three_band_path) as src:
+                transform, width, height = calculate_default_transform(
+                    src.crs, 'EPSG:32755', src.width, src.height, *src.bounds)
+                kwargs = src.meta.copy()
+                kwargs.update({
+                    'crs': 'EPSG:32755',
+                    'transform': transform,
+                    'width': width,
+                    'height': height
+                })
+
+                reprojected_path = os.path.join("/orange/ewhite/DeepForest/Wagner_Australia/three_band", basename + "_three_band_utm.tif")
+                with rio.open(reprojected_path, 'w', **kwargs) as dst:
+                    for i in range(1, src.count + 1):
+                        reproject(
+                            source=rio.band(src, i),
+                            destination=rio.band(dst, i),
+                            src_transform=src.transform,
+                            src_crs=src.crs,
+                            dst_transform=transform,
+                            dst_crs='EPSG:32755',
+                            resampling=Resampling.nearest)
+
+                png_path = os.path.join("/orange/ewhite/DeepForest/Wagner_Australia/three_band", basename + "_three_band_utm.png")        
+                # Normalize the bands to 0-255
+                bands = bands.astype('float32')
+                for i in range(bands.shape[0]):
+                    band_min, band_max = bands[i].min(), bands[i].max()
+                    bands[i] = 255 * (bands[i] - band_min) / (band_max - band_min)
+                bands = bands.astype('uint8')
+
+                # Save the normalized image as PNG
+                with rio.open(
+                    png_path,
+                    'w',
+                    driver='PNG',
+                    height=src.height,
+                    width=src.width,
+                    count=3,
+                    dtype='uint8',
+                    crs='EPSG:32755',
+                    transform=transform,
+                ) as dst:
+                    dst.write(bands[0], 1)
+                    dst.write(bands[1], 2)
+                    dst.write(bands[2], 3)
+
+        # Process the annotations
+        gdf = gpd.read_file(geopackage_path)
+        gdf["label"] = "Tree"
+
+        # project into utm S55
+        gdf = gdf.to_crs("EPSG:32755")
+        gdf["image_path"] = reprojected_path
+        annotations_df = read_file(gdf, root_dir=os.path.dirname(three_band_path))
+        annotations_df['source'] = 'Wagner et al. 2023'
+
+        # save a png without geospecific information
+        annotations_df["image_path"] = png_path
+
+        # Append to the list of annotations
+        annotations.append(annotations_df)
+
+# Concatenate all annotations into a single DataFrame
+all_annotations = gpd.GeoDataFrame(pd.concat(annotations, ignore_index=True))
+# Save to CSV
+all_annotations.to_csv(output_csv, index=False)

data_prep/collect_tasks.py

@@ -19,7 +19,8 @@
     "/orange/ewhite/DeepForest/Jansen_2023/pngs/annotations.csv",
     "/orange/ewhite/DeepForest/Troles_Bamberg/coco2048/annotations/annotations.csv",
     "/orange/ewhite/DeepForest/Cloutier2023/images/annotations.csv",
-    "/orange/ewhite/DeepForest/Firoze2023/annotations.csv"
+    "/orange/ewhite/DeepForest/Firoze2023/annotations.csv",
+    "/orange/ewhite/DeepForest/Wagner_Australia/annotations.csv"
     ]
 
 # Current errors
@@ -183,7 +184,6 @@
         shutil.copy("/orange/ewhite/DeepForest/MillionTrees/TreePolygons_v0.0/images/" + image, destination)
 
 # Write examples from the mini datasets to the MillionTrees doc folder
-
 mini_TreeBoxes_annotations.root_dir = "/orange/ewhite/DeepForest/MillionTrees/MiniTreeBoxes_v0.0/images/"
 mini_TreePoints_annotations.root_dir = "/orange/ewhite/DeepForest/MillionTrees/MiniTreePoints_v0.0/images/"
 mini_TreePolygons_annotations.root_dir = "/orange/ewhite/DeepForest/MillionTrees/MiniTreePolygons_v0.0/images/"

docs/datasets.md

@@ -98,3 +98,5 @@ https://www.mdpi.com/2504-446X/7/3/155
 
 Australia
 https://www.mdpi.com/2072-4292/16/11/1935
+
+![sample_image](public/Wagner_et_al._2023.png)