Initial commit

LICENSE

+The MIT License (MIT)
+Copyright © 2021 NILU
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

README.md

+# osmaug v0.01dev
+
+A package for:
+- getting OpenStreet Map pictures around specific locations.
+- implementing simple picture augmentation.
+
+## Install
+
+Package 'osmaug' require [geckodriver](https://github.com/mozilla/geckodriver/releases).
+
+```bash
+cd /tmp
+git clone https://git.nilu.no/aqdl/osmaug_pkg.git
+cd osmaug_pkg
+pip3 install -e .
+```
+
+## Usage
+```bash
+python osmaug/getosm.py -c config.json
+
+python osmaug/augmosm.py -c config.json
+```
+
+## Example
+```bash
+python osmaug/getosm.py -c osmaug/example/osm.json
+
+python osmaug/augmosm.py -c osmaug/example/augmosm.json
+```
+
+## Related works
+
+The package is a close adaptation of : [osm2img](https://git.nilu.no/deepair/osm2img),[osmsubsetter](https://git.nilu.no/deepair/osmsubsetter),[imgaug](https://git.nilu.no/deepair/imgaug).

osmaug/augmosm.py

+import argparse
+from typing import Dict
+import json
+import pandas as pd
+import os
+from skimage import io, img_as_float, img_as_ubyte
+from PIL import Image
+import numpy as np
+
+
+from osmaug.dictionnary.dictionnary_tf import augm_dict
+from osmaug.transformation.crop import crop
+
+def augmosm_function(params: Dict):
+    workdir = os.path.join(params['workdir'])
+    if not os.path.exists(workdir):
+        sys.exit('workdir does not exist: '+workdir)
+    resdir = os.path.join(workdir,"augm")
+    if not os.path.exists(resdir):
+        os.mkdir(resdir)
+    print("resdir: "+resdir)
+    files = pd.DataFrame.from_dict(params['files'], orient='columns')
+    for index, row in files.iterrows():
+        print("-> "+row['name'])
+        img = img_as_float(io.imread(os.path.join(workdir,row["name"])))
+        augm = augm_dict[row['augmentation']]
+        img = augm(img)
+        img = crop(img, delta=row["delta_pixel"])
+        pil_img = Image.fromarray(img_as_ubyte(img)).save(os.path.join(resdir,"aug_"+row['augmentation']+"_"+row["name"]))
+
+def main():
+    """
+    Main augmosm_function.
+    """
+    parser = argparse.ArgumentParser(description="Argument parsing for augmenting osm pictures")
+    parser.add_argument("-c", help="Config")
+    args = parser.parse_args()
+    with open(args.c,encoding='utf8') as f:
+        augmosm_config = json.load(f)
+    augmosm_function(augmosm_config)
+    print("OSM pictures augmentation completed.")
+
+if __name__ == "__main__":
+    main()

osmaug/dictionnary/dictionnary_tf.py

+from osmaug.transformation.affine import tf_from_center_random
+
+augm_dict = {
+    "affineRandom": tf_from_center_random}

osmaug/example/augmosm.json

+{
+  "files" : [
+    {
+      "name": "bygdøy_zoom_19.jpg", "augmentation" : "affineRandom", "delta_pixel" : 112
+    },
+    {
+      "name": "alna_zoom_19.jpg"  , "augmentation" : "affineRandom", "delta_pixel" : 112
+    }
+  ],
+  "workdir" : "C:/Users/jml/Documents/osm"
+}

osmaug/example/osm.json

+{
+  "points" : [
+    {
+      "name": "bygdøy", "coords" : [59.919306, 10.6965],   "conv_crs"    : "+proj=utm +zone=32V, +north +ellps=WGS84 +datum=WGS84 +units=m +no_defs", "delta_meters" : 100
+    },
+    {
+      "name": "alna"  , "coords" : [59.927689, 10.846545], "conv_crs"    : "+proj=utm +zone=32V, +north +ellps=WGS84 +datum=WGS84 +units=m +no_defs", "delta_meters" : 100
+    }
+  ],
+  "zoom" : 19,
+  "workdir" : "C:/Users/jml/Documents/",
+  "driver_bin" : "C:/Users/jml/Documents/geckodriver.exe"
+}

osmaug/getosm.py

+import argparse
+from typing import Dict
+import json
+import pandas as pd
+import os
+from osmaug.osm.map import make_bbox, load_map, map2png, png2jpg
+
+def getosm_function(params: Dict):
+    driver_bin = params['driver_bin']
+    workdir = os.path.join(params['workdir'])
+    if not os.path.exists(workdir):
+        sys.exit('workdir does not exist: '+workdir)
+    resdir = os.path.join(workdir,"osm")
+    if not os.path.exists(resdir):
+        os.mkdir(resdir)
+    print("resdir: "+resdir)
+    zoom = params['zoom']
+    points = pd.DataFrame.from_dict(params['points'], orient='columns')
+    for index, row in points.iterrows():
+        print("-> "+row['name'])
+        tmp = make_bbox(row['coords'],row['conv_crs'],row['delta_meters'])
+        file = os.path.join(resdir,row['name']+'_zoom_'+str(zoom)+'.html')
+        load_map(tmp,zoom,file)
+        tmp = map2png(file, 5, driver_bin)
+        tmp = png2jpg(tmp)
+
+def main():
+    """
+    Main get_osm_rgb function.
+    """
+    parser = argparse.ArgumentParser(description="Argument parsing for getting osm pictures")
+    parser.add_argument("-c", help="Config")
+    args = parser.parse_args()
+    with open(args.c,encoding='utf8') as f:
+        getosm_config = json.load(f)
+    getosm_function(getosm_config)
+    print("Getting OSM as RGB picture completed.")
+
+if __name__ == "__main__":
+    main()

osmaug/osm/map.py

+from pyproj import Proj
+import time
+from folium import Map
+from selenium import webdriver
+from PIL import Image
+
+# [59.92773,10.84633]
+def make_bbox(coords,crs,delta_meters):
+    """
+    Make a bbox from a location in long/lat
+    """
+    utmproj = Proj(crs)
+    # latlong2utm
+    utmx, utmy = utmproj(coords[1], coords[0])
+    # utm2latlong + delta for bbox
+    ll_lon, ll_lat = utmproj(utmx - delta_meters,utmy - delta_meters, inverse=True)
+    ur_lon, ur_lat = utmproj(utmx + delta_meters,utmy + delta_meters, inverse=True)
+    #bbox
+    res = [ll_lat,ll_lon,ur_lat,ur_lon]
+    return(res)
+
+# bbox=[59.8,10.7, 60.0, 10.8]
+def load_map(bbox,zoom,file):
+    """
+    Load OSM leaflet and save it.
+    """
+    # lat/long
+    center=[bbox[0]+(bbox[2]-bbox[0])/2,bbox[1]+(bbox[3]-bbox[1])/2]
+    bounds=[[bbox[0],bbox[1]],[bbox[2],bbox[3]]]
+    # get map
+    m = Map(center, control_scale=True, zoom_start=zoom)
+    m.fit_bounds(bounds, max_zoom=zoom)
+    m.save(file)
+
+def map2png(file, delay, driver_bin):
+    """
+    Convert .html map tile to .png and save it.
+    """
+    driver = webdriver.Firefox(executable_path=driver_bin)
+    driver.get('file://{}'.format(file))
+    # Give the map tiles some time to load
+    time.sleep(delay)
+    print(file)
+    driver.save_screenshot(file.replace('.html', '.png'))
+    driver.quit()
+    return(file.replace('.html', '.png'))
+
+def png2jpg(file):
+    """
+    Convert image file from .png to .jpg and save it.
+    """
+    img = Image.open(file).convert('RGB')
+    img.save(file.replace('.png','.jpg'))
+    return(file.replace('.png','.jpg'))

osmaug/transformation/affine.py

+
+import random
+import numpy as np
+from skimage import io, transform, util
+
+
+def tf_from_center_random(img: np.ndarray):
+    """
+    Transform an image around its center with random rotation, random shear and reflection
+    """
+    # For a transformatino around the center of the image
+    # one compose a translation to change the origin, a transformatino, and finally the inverse of the first translation.
+    # https://scikit-image.org/docs/stable/auto_examples/transform/plot_transform_types.html#sphx-glr-auto-examples-transform-plot-transform-types-py
+    shift = transform.EuclideanTransform(translation=-np.array([img.shape[1],img.shape[0]]) / 2)
+    # random rotation between -pi and pi
+    rotation = transform.EuclideanTransform(rotation=random.uniform(-np.pi, np.pi))
+    # shear parallel to the x-axis.
+    # https://en.wikipedia.org/wiki/Transformation_matrix
+    # A shear parallel to the y axis has x' = x + ky and y' = y. Written in matrix form.
+    # Random shear between 0.1 and 0.3
+    shear_x = np.array([[1, random.uniform(0.1, 0.3), 0],
+                          [0, 1, 0],
+                          [0, 0, 1]])
+    # shear parallel to the x-axis:
+    #A shear parallel to the y axis has x' = x and y' = kx + y. Written in matrix form.
+    # Random shear between 0.1 and 0.3
+    shear_y = np.array([[1, 0, 0],
+                         [random.uniform(0.1, 0.3), 1, 0],
+                         [0, 0, 1]])
+
+    # reflection if true or none otherwise
+    refl_bool = random.sample([True,False],1)
+    if refl_bool[0]:
+        reflection = np.array([[1, 0, 0],
+                               [0,-1, 0],
+                               [0, 0, 1]])
+    else:
+        reflection = np.array([[1, 0, 0],
+                               [0, 1, 0],
+                               [0, 0, 1]])
+    # transformation matrix
+    matrix = np.linalg.inv(shift.params) @ reflection @ shear_x @ shear_y @ rotation.params @ shift.params
+    tform = transform.EuclideanTransform(matrix)
+    tf_img = transform.warp(img, tform.inverse)
+    return(tf_img)

osmaug/transformation/crop.py

+import numpy as np
+
+def crop(img: np.ndarray, delta: int=112):
+    h, w, _ = img.shape
+    cx = int(w / 2)
+    cy = int(h / 2)
+    x1 = cx - delta
+    x2 = cx + delta
+    y1 = cy - delta
+    y2 = cy + delta
+    return img[y1:y2, x1:x2, :]

requirements.txt

+scikit-learn~=0.24.1
+pandas~=1.2.4
+numpy~=1.20.2
+argparse~=1.1
+pyproj~=3.0.1
+folium~=0.12.1
+selenium~=3.141.0
+scikit-image~=0.18.1

setup.py

+from setuptools import setup
+import os
+
+with open('README.md', 'r') as fh:
+    long_description = fh.read()
+
+library_folder = os.path.dirname(os.path.realpath(__file__))
+
+requirementPath = f'{library_folder}/requirements.txt'
+install_requires = []
+if os.path.isfile(requirementPath):
+    with open(requirementPath) as f:
+        install_requires = f.read().splitlines()
+
+dev_requirements = []
+
+setup(
+    name='osmaug',
+    version='0.01dev',
+    author='Jean-Marie Lepioufle, Islen Vallejo',
+    author_email='jml@nilu.no, iv@nilu.no',
+    packages=[
+        'osmaug',
+        'osmaug.osm',
+        'osmaug.transformation'],
+    license='MIT + Copyright NILU',
+    description='A package for getting OpenStreet Map pictures and implementing augmentation.',
+    long_description = long_description,
+    url='https://git.nilu.no/aqdl/osmaug',
+    python_requires='>=3.9',
+    install_requires=install_requires,
+    extras_require={
+        'dev': dev_requirements})