From 743c7d548fb3da7b76aa50d66d1f6a56ca62eca6 Mon Sep 17 00:00:00 2001
From: Ruben van de Ven <git@rubenvandeven.com>
Date: Thu, 12 Jan 2023 16:49:51 +0100
Subject: [PATCH] Tool for extracing labels from the VLoD image dataset

---
 build_dataset_labels.py | 54 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 54 insertions(+)
 create mode 100644 build_dataset_labels.py

diff --git a/build_dataset_labels.py b/build_dataset_labels.py
new file mode 100644
index 0000000..43a8405
--- /dev/null
+++ b/build_dataset_labels.py
@@ -0,0 +1,54 @@
+import os
+from pathlib import Path
+from typing import Optional
+from PIL import Image
+
+Image.init() # required to initialise PIL.Image.EXTENSION
+
+def extract_labels(f: Path) -> dict:
+    # get the labels for the image path
+    return {
+        "arrondisement": int(str(f).split(' ')[0][:-1]),
+        "street": str(f).split('/')[1].split(' ')[0],
+    }
+
+def is_image_ext(f: Path) -> bool:
+    return f.suffix.lower() in Image.EXTENSION
+
+def open_image_folder(source_dir, *, max_images: Optional[int] = None):
+    image_files = [f for f in sorted(Path(source_dir).rglob('*')) if is_image_ext(f) and os.path.isfile(f)]
+
+    labeled_images = [{
+        "abspath": f.resolve(),
+        "relpath": f.relative_to(source_dir),
+        "labels":  extract_labels(f.relative_to(source_dir))
+        } for f in image_files]
+
+    return labeled_images
+
+def filter_by_label(image_files: list, label, value):
+    return list(filter(lambda i: i['labels'][label] == value, image_files))
+
+def aggregate_labels(image_files) -> dict[str, list]:
+    labels = {}
+    for f in image_files:
+        for label, value in f['labels'].items():
+            if label not in labels:
+                labels[label] = []
+            if value not in labels[label]:
+                labels[label].append(value)
+
+    for label, values in labels.items():
+        labels[label] = sorted(values)
+
+    return labels    
+
+def print_stats(image_files, labels):
+    for label, values in labels.items():
+        print(label)
+        for value in values:
+            print(f" - {value}: {len(filter_by_label(image_files, label, value))}")
+
+image_files = open_image_folder("../VLoD")
+labels = aggregate_labels(image_files)
+print_stats(image_files, labels)
\ No newline at end of file