New flavia images dataset

README.md

@@ -25,3 +25,11 @@ For instance, [Stanford bunny LICENSE](points/bunny/LICENSE) applies to [Stanfor
 | Model Name | Source | Description                                                                                                                                                                           |
 |------------|--------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
 | [Topological Wheels](timeseries/topological_wheels/tw-data-annotated.png) | [DataShape](https://www.inria.fr/fr/datashape) | 10 CSV files where each file has 64 timeseries over 10000 timestamps each, and a $y$ ground truth indicating anomalous regime. See [README](timeseries/topological_wheels/README.md). |
+
+## Images
+
+| Model Name | Source | Description                                                                                                                                                                           |
+|------------|--------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| [Flavia leaves](images/flavia/flavia_convexity.csv.gz) | [Flavia](https://flavia.sourceforge.net/)[^1] | Compressed CSV file with 1907 flattened black & white images (900 first values) and the image convexity value (last column value). |
+
+[^1]: *Stephen Gang Wu, Forrest Sheng Bao, Eric You Xu, Yu-Xuan Wang, Yi-Fan Chang and Chiao-Liang Shiang, A Leaf Recognition Algorithm for Plant classification Using Probabilistic Neural Network, IEEE 7th International Symposium on Signal Processing and Information Technology, Dec. 2007, Cario, Egypt*

images/flavia/data_selecta.py

@@ -0,0 +1,56 @@
+# Based on https://github.com/renata-turkes/turkevs2022on/blob/main/SRC/data_construction.py
+# Author: Renata Turkeš
+# This file is part of the Gudhi Library - https://gudhi.inria.fr/ -
+# which is released under MIT.
+# Author(s):       Sean Bontemps, Vincent Rouvreau
+#
+# Copyright (C) 2024 Inria
+#
+# Modification(s):
+#   - YYYY/MM Author: Description of the modification
+
+import argparse
+import os
+import numpy as np
+from PIL import Image
+from skimage.morphology import convex_hull_image
+
+#
+# This program requires to load data from https://sourceforge.net/projects/flavia/files/Leaf%20Image%20Dataset/1.0/Leaves.tar.bz2/download
+# This dataset is based on the paper A Leaf Recognition Algorithm for Plant classification Using Probabilistic Neural Network, by
+# Stephen Gang Wu, Forrest Sheng Bao, Eric You Xu, Yu-Xuan Wang, Yi-Fan Chang and Qiao-Liang Xiang, published at
+# IEEE 7th International Symposium on Signal Processing and Information Technology, Dec. 2007.
+#
+# Unzip the data and copy data_selecta.py in a same directory data folder
+# Launch 'python data_selecta.py -s /my/path/to/flavia'
+# Generated with Python 3.8 / Pandas 1.1.0
+#
+
+parser = argparse.ArgumentParser(
+    description="Reads all the images from a source_directory, converts them to (image_size x image_size) black &"
+    " white images, computes their convexity. The final numpy ndarray (number of images x (image_size x image_size + 1)"
+    " is saved in a destination_file."
+)
+parser.add_argument("-s", "--source_directory", type=str, default="flavia")
+parser.add_argument("-d", "--destination_file", type=str, default="flavia_convexity.npy")
+parser.add_argument("-i", "--image_size", type=int, default=30)
+
+args = parser.parse_args()
+
+data = []
+for image in sorted(os.listdir(args.source_directory)):
+    if image.endswith('.jpg'):
+        file_name = args.source_directory + "/" + image
+        # print(file_name)
+        img = Image.open(args.source_directory + "/" + image)
+        # Convert image in gray scale and resize
+        img = img.convert('L').resize((args.image_size,args.image_size), Image.LANCZOS)
+        # Black & white image at a specific threshold
+        img_bw = img < 0.9 * np.max(img)
+        img_bw_ch = convex_hull_image(img_bw, offset_coordinates = False)
+        convexity = np.sum(img_bw) / np.sum(img_bw_ch)
+        data_line = np.append(img_bw.astype(float).reshape(args.image_size * args.image_size), convexity)
+        data.append(data_line)
+
+data = np.asarray(data)
+np.save(args.destination_file, data, allow_pickle=False)

images/flavia/flavia.LICENSE

@@ -0,0 +1,11 @@
+The original dataset comes from
+https://sourceforge.net/projects/flavia/files/Leaf%20Image%20Dataset/1.0/Leaves.tar.bz2/download
+
+This dataset is based on the paper A Leaf Recognition Algorithm for Plant classification Using Probabilistic Neural
+Network, by Stephen Gang Wu, Forrest Sheng Bao, Eric You Xu, Yu-Xuan Wang, Yi-Fan Chang and Qiao-Liang Xiang, published
+at IEEE 7th International Symposium on Signal Processing and Information Technology, Dec. 2007.
+
+Please read carefully the terms of use of this dataset (https://flavia.sourceforge.net/) before using it.
+
+The images have been black and whited, resized in 30 x 30 and then flattened. The last column of the dataset is the
+convexity value. Please refer to data_selecta.py for more details.