ICPC faces

Automated face recognition system with badge detection

Watch deployed at https://news.icpc.global/gallery/

Contents

1. How to use
2. ICPC Photo Tagging process
3. Known issues

How to use

1. Automated face recognition and badge recognition system.
2. Multiple scripts for easy tag manipulation.

ICPC Photo Tagging process

Goal is to identify as many people on the photos as possible, and for each person mark the corresponsing Badge name for further use in the ICPC Gallery news.icpc.global/gallery.

Sources of information are:

• list of all ICPC attendees with their role, institution (for teams ++) and Badge Name. Notice, that Badge Name may be different from passport first namd and last name. Within ICPC we try to address people as their Badge Name.
• digikam software that identifies faces on images, finds similar faces and helps organize the photos, add tags and so on
• multiple command line tools for different tagging formats interoperability
• Azat badge-to-face detection system for Team Photos.

digikam

The software we suggest to use for photo management is digikam. When setting up digikam it is important to

• select a folder, where the all photos will be stored
• enable option store tags in photos

Pre-select training data

Core organizer team doesn't change much between years. Prepare a separate folder with core org group photos tagged from last years to help the digikam identify people.

Team Photos -- Azat

Another main source of information about specific WF attendees are Team Photos. Every team member on the team photo is expected to have a visible badge with Badge name and university name printed on them. If you provide Azat software with the list of all team members and their correct badge names, the tool will identify almost all faces and corresponding people. It will write the result in a human-readable file for manual check.

Confirmed informationa from file can be then be burned into tags of these photos. Tags will work on both digiKam and Picasa face detection formats.

Adding new events

Photos from every event should be placed in a separate folder. It is also possible to put photos from different photographers into different folders for easier author tagging.

Every time you add a folder to the root digikam photo folder, you will need to refresh the list of files in digikam. Then run the people -> detect faces (skip already scanned photos) and people - recognize faces (skip already scanned photos). You will see a list of unknown or unconfirmed recognized faces that need to be confirmed or rejected manually.

When all face tagging for an event is ready, create a new tag with event$ prefix, for example event$Opening Ceremony. Select all new photos in the folder, right-click, select Assign tag and select the appropriate tag. Then it will be applied to all selected photos.

Each photo needs to be tagged with a year tag: album$2021.

Then you will need to run a tool convert_digiKam_to_picasa.py to embed photo tags from digikam into appropriate for icpc picasa tagging format.

Editing tags on Flickr

If a wrong tag was already uploaded to the flickr, they can still be easily edited through flickr interface. Make sure to include quotation marks.

Known issues

• how to edit photo description to include photographer name?
• university names with special symbols & , can be processed incorrectly