whereispatrick.at

Downloading geo-tagged photos from Google Drive

My father takes geo-tagged photos (of some pretty amazing things) with his regular cellphone. These photos get uploaded automatically to his Google Photos account whenever he connects to wifi. However, Google Photos does not have an API. Initially, it seemed that there was no easy way to programatically access these pictures. Here is the work-around.

Start by setting up the Google drive from the same account to have a folder pointing towards the Google Photos. This is a setting in Google Drive: go to the Google Photos tab in the Google Drive, click on the settings icon in the top right corner. When you scroll down, you will see: Create a Google Photos folder Automatically put your Google Photos into a folder in My Drive. Make sure this is set. You will then see a folder inside Google Drive named Google Photos. Now, because Google Drive does have an API, you can access the photos through that folder.

The Google Drive API Python example is a great place to start, specifically to register the app and get the client ID and a client secret. More details on the authorization can be found here: https://developers.google.com/drive/v2/web/about-auth. Here are a few tweaks and tricks that I had to do for the whereispatrick project:

• First, since this an app that will download photos from the drive, the scope needs to adjusted. In this case, I have just set the scope to https://www.googleapis.com/auth/drive. This is full access. Writing this, I realize that this is likely too broad of an authorization and created a GitHub issue to look into testing the https://www.googleapis.com/auth/drive.readonly scope.

• Second, since this app runs inside a Docker, we need to make sure that we can get the Oauth through the command line, this is done using the --noauth_local_webserver option.

In this case, I only had to get the authorization once. The resulting credentials are saved locally and shared with the appropriate docker containers. Once the authorization is in order, the photos can be accessed using the apiclient module, see images.py here: https://github.com/Freija/whereispatrick/blob/master/images/images.py. In that code, the images are downloaded and converted to a smaller thumbnail format using PIL. The full-sized images are then deleted. The GPS coordinates are part of the EXIF information. This is accessed using the piexif module. The piece of info we want is in the GPS section of the EXIF. This can be different for different camera models. In this case, this is the format of that info (example data):

{
  0: (2, 2, 0, 0),                          GPS tag version
  1: 'S',                                   North or South latitude
  2: ((11, 1), (52, 1), (261135, 10000)),   latitude
  3: 'W',                                   East or West longitude
  4: ((75, 1), (17, 1), (398785, 10000)),   longitude
  5: 0,                                     altitude reference level
  6: (3422, 1),                             altitude
  7: ((15, 1), (29, 1), (10, 1)),           time
  27: some crazy stuff,                     undefined data (?)
  29: '2017:07:12'                          date
}

The thumbnails are saved to disk and served from a separate nginx docker. The coordinates of the photo location and image name is saved to a local CSV file in the following format:

1,  IMG_20170716_160455.jpg,  -10.871571527777778,  -75.08304594444444, 582,2017-07-16 21:04:53
1,  IMG_20170716_141302.jpg,  -10.852666833333332,  -75.10791777777777, 891,2017-07-16 19:06:39
1,  IMG_20170716_125418.jpg,  -10.855616555555555,  -75.11288449999999, 766,2017-07-16 17:54:16
1,  IMG_20170716_160026.jpg,  -10.869841555555556,  -75.08080291666667, 584,2017-07-16 21:00:24
1,  IMG_20170716_113522.jpg,  -10.946722972222222,  -75.26346586111111, 723,2017-07-16 16:35:21

The first column tells us if we should plot these coordinates; images that do not have valid coordinates will not be plotted. The next columns are the image name, latitude, longitude, altitude and date. This file is also used at the time of the Google Drive download to check if we already have downloaded and processed a file. This last check is needed since there seems to be no clear order in which photos are uploaded from my father’s phone. In fact, some photos might appear many days after photos from the same date.

The Google Drive API allows for 1 billion requests per day, more than enough for me!

Partial screenshot of the Google Drive API dashboard showing the usage versus time.

1, "[-10.934536916666667, -75.19217680555556]",
   "[
     ['IMG_20170716_164718.jpg', -10.934536916666667, -75.19217680555556, 413, '2017-07-16 21:47:15']
    ]"
2, "[-11.033253423698437, -75.31624411811225]",
   "[
     ['IMG_20170716_103053.jpg', -11.033305166666667, -75.31606291666667, 735, '2017-07-16 15:30:52'],
     ['IMG_20170716_103012.jpg', -11.033273694444445, -75.31622313888889, 735, '2017-07-16 15:30:10'],
     ['IMG_20170716_102536.jpg', -11.033129666666667, -75.3166275, 906, '2017-07-16 15:25:34']
    ]"

<div id="map"></div>
<script src="/index.js"></script>
<script async defer src="https://maps.googleapis.com/maps/api/js?key=<pass your key here>&callback=initMap"></script>

version: '2'
services:
  app:
    build: app
    image: "freija/whereispatrick"
    volumes:
     - "./data:/data"
     - "./config:/config"
    environment:
     - "STATIC_URL_BASE=http://localhost:8000"
  static:
    image: "nginx:alpine"
    volumes:
      - "./data:/usr/share/nginx/html:ro"
  images:
    build: images
    image: "freija/images"
    volumes:
      - "./data:/data"