See shmd.org to learn what SHMD is all about.
One of the many groups of participating volunteers is the photo crew. Numerous photographers take thousands of photos of every aspect of the dance. These are funneled back to a central server, where they are distributed to printing stations for purchase by attendees. All proceeds go straight to the SHMD fund pool.
In 2016, with little notice, the photo crew was informed that their usual set up would no longer be possible to support. Due to lack of sufficient power supply for all the printing and networking hardware, the ordering and printing stations would have to be split into two separate locations. This implied the need the create some mechanism to relay print orders from the former to the latter. This application was hastily thrown together in a weekend in an attempt to address that requirement.
After great success in 2016, the app was updated in 2017 to perform facial recognition via the Amazon Rekognition service. Armed with student metadata and high quality individual reference portraits from the yearbook vendor, the Amazon Rekognition engine can be seeded with baseline images for all students. This process creates a unique FaceId for each student, which is kept in a SQLite databse. As event photos pour in, they may be passed to Rekognition to identify known faces. The results are returned as FaceIds, which are then cross referenced to student name, and stored in the database by photo. This allows for very quick lookup of all photos that a given student appears in.
This is a PHP app, just point your web server's document root to the public
subdir. Make sure that the orders and orders/archive directories are
writable by the web server user. Run composer install to fetch dependencies.
In order to use the facial recognition features, you'll need an Amazon AWS
account. Copy config.json.dist to config.json and edit as indicated.
The app supports an arbitrary number of galleries. The intent is to separate
photos into groups by event and/or time, so that patrons will have an easier
time finding photos to purchase. To create a gallery, create a subdirectory
under the staging directory. Use lowercase names with no spaces or special
characters. The directory name will be used as the URL slug for the gallery.
Create another subdirectory with the same name under the public/photos
directory. Make sure it's readable by the web server user. If you would like
the app to show a more descriptive name for the gallery, you may create a text
file public/photos/<gallery>/title containing the gallery title. You may also
create public/photos/<gallery>/description with a short description of the
gallery contents and public/photos/<gallery>/highlight with the name of the
photo to use for the index page highlight.
As raw photos from the photographers are acquired, they should be copied into
the appropriate staging/<gallery> directory. This is the only location that
photos should be manually placed. Note that these files can be quite large and
will never be served raw. At any time, you may run the bin/resize.sh script.
This will scan for newly added photos in staging, resize them to something
more appropriate for the web, and copy them to public/photos/<gallery>.
Photos will not appear in the app until after they have been so processed.
Note that the files in public/photos/<gallery> must be readable by the web
server process user.
To seed the facial recognition database, run the script bin/indexPhotos.php
which will import your yearbook portraits into your Rekognition collection.
To perform facial recognition on photos, run the bin/identifyPhotos.php
script. This can be run at any time, it will not interfere with app usage. The
process will update the SQLite database and enable searches by name in the app.
The app supports printing of order slips via a USB ESC/POS receipt printer. The printer must be connected to the same machine running the server.
The machine acting as server should be located physically close to the printing
stations so that the printing personnel have access to the order slips. Once
photos are processed with the bin/resize.sh script, they will appear in the
app and be available for purchase. Client machines need only a webserver.
Laptops, Chromebooks, iPads, and even SmartBoards work well. Dance patrons
and/or sales agents can use the app to browse the galleries, and select
individual photos for purchase. Submitting the order form will create a new
order. If a printer is connected, an order slip will immediately print. Workers
at the printing stations can grab slips as they appear, print the requested
photos, and run the order out to the patron. Alternatively, print workers may
browse the /orders URL to watch for incoming orders. Once fulfilled, they may
click the Complete button, which will archive the order and remove it from the
order page so that it does not get printed twice.
Pending orders are stored in the orders subdirectory. There will be one file
per order, named with a unique hash, and formatted in plain text JSON. Once an
order is fulfilled, the file will be moved to the orders/archive directory.
These files can be easily parsed for statistics once the event is complete.
Note that if order slips are sent to a receipt printer, then the orders are
automatically archived and do not need to be fulfilled from the /orders URL.