show and tell for hackers in DC.

Round 35: Omniscient Omnibus

09 Aug 2016

Shannon Turner - @svthmc

BudgieCam: I used a Raspberry Pi and camera to get pictures of my pretty bird whenever I send a text message!


Bird is eating but looks like Bernie Sanders.

Moving parts: Twilio, Django, Apache on Raspberry Pi. Hard part: Port forwarding on the Pi. We all need to get birds and tweet pictures of them to @svthmc

Cost: $30 (Pi) + $25 (Pi camera) + Cable + Twilio account

BCARV: BudgieCam Also Records Videos

Scraper to collect all the images

Password protected to avoid snoopers! Long cable created noise in the photos. Also may try increasing exposure length

Christian Flores - @chrifs

Built an iOS app that shows nearby DC metro, bus, circulator, streetcar + bikeshare info, maps, walking directions for stops, and alerts.

First time building a mobile app, made it in Swift

Goals in creating it: Use public APIs Reduce backend work (no servers) simplifying the user experience

See bus stops within 3/4 miles of your location Tells you all the buses at this stop and when they are arriving Lets you see routes and maps of the buses, overlaid on Google map with lots of landmarks!

Also has Metro – shows the entrances and Bikeshare, where the bikeshare stops are

Transit APIs: DC Metro (Metrorail) & Metrobus DC Circulator Capitol Bikeshare

Used Inkscape (free), GIMP (free), and Sketch ($99)

Available in iOS only Most difficult part was learning closures Swift is much friendlier to learn than Objective-C used to log app crashes

Jess Garson - @jessicagarson

Light up shoes part 2! In which we are promised a LIVE DEMO Uses the Adafruit Gemma – an improvement from last month’s Arduino Uno Uses C++ (on the Arduino) Soldering, sewing, and working together with other Hack and Tellies!

There are two shoes now! We’re getting very close to being able to hold a dance party! A brief dance party was held.

Using the code at over It doesn’t have to just be EDM!

Ben Klemens - @b__k

The pop charts are a weird and surprising place. @popchartbot posts a lyric/day from a pop song.

Ben wrote a bot that tweets one lyric from a pop chart topping song – they’re often bizarre and out of context, very strange!

Seems weird because all of these songs are theoretically mainstream songs, right? But they’re very strange.

We start with a list of all of the pop chart songs: The Whitburn Project, 120 years of music chart history Used because it was good enough – though there are egregious errors in all of them! But this site had no rate limiting and consistent URLs

The lyrics are artisinally hand-picked by real live human people

It’s fun and a bit of mystery that the artist and song title isn’t mentioned!

140 characters limits how much info we can cram into a tweet

Mike Mahony

In response to DC police announcing that they would no longer tweet crime alerts, I created a twitter bot with Python and Flask that tweets alerts from an official RSS feed (which is public). Because twitter’s short format is not optimal for the often long alerts, I expanded the flask app to permanently host alerts in their full form.

The twitter bot was released as @CrimeAlertsDC, but is now @AlertDCio

Originally only a twitter bot that looked at the RSS feed but expanded to a webapp, hence the re-branding. Now permanently hosts the complete alerts at Alerts are typically much longer than 140 characters so it’s much more useful to be able to link to the full details Written in Python and Flask; uses Tweepy (Twitter API wrapper for Python); Postgres and SQLAlchemy; and BeautifulSoup

Possible expansions: filter by area

Analytics: Power Outages get 10-20x as many impressions / clicks / views / etc than any other kind of alert

Unusual events drive traffic, like the person who stabbed someone but got away on a segway

Don’t use this to discern any conclusions about crime in the city!

Eric Schles - @EricSchles

This talk will walk through a tool used to analyze sex trafficking Trip to India as a child; saw a human slave - broken limbs so people feel sorry for you and give you money Scraping Some percentage of escorts are human trafficked Analyze emotional valance of posts to get an idea of the relationship between the victims and the traffickers

Travis Hoppe - @metasemantic

Watch the birth of baby networks! Training NN to visualize the patterns made as the network learns.

Deep learning and neural nets - learning what the connections mean Shallow learning versus deep learning Shallow network is on the right track for a while, but with high uncertainty .. Deep networks are “very confident and very wrong” for a while Tried different activation functions How does the neural network learn? Visualizing how the neural networks are figuring out what shapes are Given the X,Y coordinates, can you figure out what it looks like? More importantly, not can it learn it perfectly, but how does it learn, and what does it look like to the neural network as it runs

Given X,Y coordinates of the actual image, with the classes as colors

Most helpful from a pedagogical standpoint, to understand the noticeable difference between how shallow and deep neural networks learn

Source code available at

Check out Aaron Schumacher’s Intro to Tensor Flow!