Mapillary API v4 Toolset

Useful tools for interacting with Mapillary API v4

This repository includes a set of easy-to-use Python scripts that interact with Mapillary API v4 and can do a set of basic and useful operations.

For now the current functionalities developed include:

1. Fetching image sequences of a user or organization
2. Downloading image ids of user sequences based on the output of the previous (1)
3. Downloading image metadata of images based on the output of the previous (2)
4. Download the actual images on the given quality (configuratble through the settings file)
5. Fetch image detections. The set of detections to fetch is configurable through the settings file
6. Visualize detections on top of the downloaded images

The settings.py file is defining a dictionary of settings, and each script is reading the ones it needs. Thus, to modify the inputs/settings for your needs (like changing the Mapillary token, the bounding box, etc.) you can modify the settings.py file. More information on the settings is shared later on.

Script to get sequences: get_sequences.py

This script is fetching all the sequence IDs for a given organization ID (or user ID) in a given bounding box. It generates a csv file that includes the sequence ids of the organization given in the bounding box specified.

In particular, each row will include 2 items.

1. org_id - the organization id (or user id) that owns the sequence
2. seq_id - the sequence id

Example Output:

org_id,seq_id
803468386961584,zK7pW0F1mgnaLiZbDqOjRI
803468386961584,Mr1jEG5HXBJxLfUS9Wtlp2
803468386961584,f72dwjtsgO6NQWEecSr5vU

In order for it to run properly, you have to make sure that the following settings in settings.py have been properly set:

• token - The Mapillary client access token that is required to query the API. It defaults to a random one, so you have to change it to your own. Let me know if you need help finding it!
• org_id - The organization id (or user id) that will own the sequences. You should be able to find yours through your Mapillary profile. Let me know though if you need help. The default value is random, thus will probably fetch no results
• min_zoom - The lowest zoom level that will be considered while searching for sequences. Defaults to 1. I don't think you should change it
• max_zoom - The maximum zoom level that will be considered while searching. Deafults to 14. I don't think you should change it
• output_folder - The directory where the outputs will be stored. It is considered to exist inside the current directory. If it doesn't exist, it will be created. The default is outputs.
• seq_out_file - The name of the output file that will hold the sequence ids. It defaults to seq_dataset.csv
• The bounding box settings dictionary (AOI_bbox). The default values include Cyprus
  • ll_lat - Lower left latitude of our bounding box
  • ll_lon - Lower left longitude of our bounding box
  • ur_lat - Upper right latitude of our bounding box
  • ur_lon - Upper right longitude of our bounding box

Script to get image IDs of sequences get_images.py

This script is using the output of get_sequences.py as its input. It goes through the sequence IDs in that file, and retrieves the image IDs that belong to each. It generates a csv file that includes the pairs of sequence IDs and image IDs.

In particular, each row will include 2 items.

1. seq_id - a sequence id
2. img_id - id of an image belonging to that sequence

Example Output:

seq_id,img_id
SWxf3IwpsCM9GKBaetvkAy,1463097064038568
SWxf3IwpsCM9GKBaetvkAy,419363189104763
SWxf3IwpsCM9GKBaetvkAy,235997291422947

In order for it to run properly, you have to make sure that the following settings in settings.py have been properly set:

• token - The Mapillary client access token that is required to query the API. It defaults to a random one, so you have to change it to your own. Let me know if you need help finding it!
• output_folder - The directory where the outputs will be stored. It is considered to exist inside the current directory. If it doesn't exist, it will be created. The default is outputs
• seq_out_file - The output file of the get_sequences.py script. It will be used as input. You should not change it, unless you changed it previously, to run the get_sequences.py script.
• img_out_file - The name of the output file that will hold the sequence and image id pairs. It defaults to img_dataset.csv

Script to get image details (image metadata) get_images_metadata.py

This script is using the output of get_images.py as its input. It goes through the image IDs in that file, and retrieves the image metadata that are configured in the settings.py file to be fetched. It generates a csv dataset file that includes the following (with the current configuration):

• seq_id - The sequence ID
• img_id - The image ID
• captured_at - The time of image capture (in miliseconds since unix epoch)
• compass_angle - The compass angle of the image
• thumb_256_url - The download url of the image in the lowest possible quality (for storage purposes), which is a 256-pixel wide thumbnail. You can easily configure the script to get a better quality image, don't worry!
• width - The width (in pixels) of the original uploaded image. I am including this as for my purposes it helps me understand which camera it was captured with.

Example Output:

seq_id,img_id,captured_at,compass_angle,thumb_256_url,width
SWxf3IwpsCM9GKBaetvkAy,935274303987496,1624790409193,184.43365478516,https://scontent.fath3-3.fna.fbcdn.net/m1/v/t6/An8SGFsS9Qc6zc_oGeGBjCnvE_YRykhTtQAtDDPEAT8gRi9P_e0WKmXL6ArVzayE4qEZ--YjlNEXluClT5nKoiXnfb2AvRW02kyuzZib_7EFNgvn2YLRw7hHcdbbe6zMzYaPHT_sIUMwA8U-Lx4xLg?stp=s256x144&ccb=10-5&oh=00_AT_vXUYInVUt62p-jt4cha31oXsYlI1RqeMZWSzhsVl06Q&oe=62A68E04&_nc_sid=122ab1,2560
SWxf3IwpsCM9GKBaetvkAy,571180097599390,1624790410195,184.89834594727,https://scontent.fath3-3.fna.fbcdn.net/m1/v/t6/An9yfyxI-EoHFSYBt7f1pjVbdpFHyCKpnWzYxgLh7V4AeycSGCw3s00RpbfxuGJU-2vvGWnxGQtzN9x9tLF81e3ix806wWKmVpcnTFCDmOwMV1B0UfKMX4vimDhiu8vekk9G8FziGz3qM8C1LQjxFg?stp=s256x144&ccb=10-5&oh=00_AT8XIp6IMHZbCpN1mEvyA1vJZ2Uq9RbaSk7l4GRRdSpOuQ&oe=62A7DBBC&_nc_sid=122ab1,2560
SWxf3IwpsCM9GKBaetvkAy,323654212518788,1624790412197,184.50595092773,https://scontent.fath3-3.fna.fbcdn.net/m1/v/t6/An-BfmG9eVwf0v6zIZQG6T3Ff-2paZ0FajpslmD0Iww7sFns4x16VOB817jsYgCz4mxgLQ4lYx__sIBDwBXFZ7X6fvjemVkGkFfEGVAYSXIleN330nhUGPVdkXEp8dMKSz09QTFA4LzQNNRz7LKyoQ?stp=s256x144&ccb=10-5&oh=00_AT-eZtxkMFIxl_KhY7f2Ot0g0N0QydiMV-F3LUo0VkRgCg&oe=62A78004&_nc_sid=122ab1,2560

In order for it to run properly, you have to make sure that the following settings in settings.py have been properly set:

• token - The Mapillary client access token that is required to query the API. It defaults to a random one, so you have to change it to your own. Let me know if you need help finding it!
• output_folder - The directory where the outputs will be stored. It is considered to exist inside the current directory. If it doesn't exist, it will be created. The default is outputs
• img_out_file - The name of the output file of the get_images.py script. This script uses that file as input. It defaults to img_dataset.csv
• img_meta_out_file - The name of the output csv dataset file. It will hold all the metadata information for each image. It defaults to img_meta_dataset.csv
• graph_endpoint - The Mapillary API v4 graph endpoint. You don't need to change that. The default is the correct one.
• img_request_batch_size - The batch size that will be used to simultaneously fetch results for images in a single request. Defaults to 50. I think you should not change that.
• image_metadata_fields - The fields that are requested for each image. Change this if you need more/less info. By default it will fetch the columns described above and displayed in the example output.

Script to download images download_images.py

This script is using the output of get_images_metadata.py as its input. It downloads the images included in that file and stores them in a specific directory, that defaults to ./outputs/downloaded_images/. The images are stored with the filename following the convention <imageID_quality.jpg>. Thus, an image with id 12345 and quality 256, will be stored as 12345_256.jpg.

Note that before attempting to download, the script will check if that particular image in that particular quality already exists. If it does, it will skip it. This means that if you start downloading lots of images, and for some reason the script stops, you can then re-run it and it will automatically pick up from where it left.

In order for it to run properly, you have to make sure that the following settings in settings.py have been properly set:

• output_folder - The directory where the outputs will be stored. It is considered to exist inside the current directory. If it doesn't exist, it will be created. The default is outputs
• image_downloads_folder - The folder name inside the outputs folder to store the actual images (or image thumbnails). Defaults to downloaded_images
• img_download_quality - The quality of image download. It defaults to 256. Possible values are:
  • 256 - For a thumbnail of width 256 pixels (and adjusted height) (worst quality - smallest size)
  • 1024 - For a thumbnail of width 1024 pixels (and adjusted height)
  • 2048 - For a thumbnail of width 2048 pixels (and adjusted height)
  • original - For the original image (best quality - largest size)

Thus a use case scenario, is for an organization or user that wants to receive their images. They will

1. Run the get_sequences.py script using their user_id or organization_id and specifying their bounding box of interest
   • The result of the above execution is stored in the outputs folder (or however the user selects to change it) as _seqdataset.csv
2. Run the get_images.py script, which will read the sequences from the _seqdataset.csv and fetch the image ids.
   • The result of the above execution will be stored as _imgdataset.csv in the outputs folder
3. Run the get_images_metadata.py script, which will read the image IDs from _imgdataset.csv and fetch the metadata of these imges.
   • The result of the above execution will be stored as _img_metadataset.csv
4. Run the download_images.py script, which will download all the images, which are present in the _img_metadataset.csv file
   • The images will be downloaded according to the configured quality (see instructions above) in the directory outputs/downloaded_images

This README file will be updated soon with more details. Let me know if you need specific instructions!

Also, the repo will be updated with more functionality. Next step is to incorporate detections.