PyCon 2015 Proposals

Deadlines for PyCon 2K15 closed yesterday, and this year I submitted two different talks in a shotgun-blast style attempt to get accepted. Last year I was kicking myself for proposing a talk that committed myself to a ton of work, but in the end it was probably worth it. You can see the talk from 2014 on YouTube. My two proposals this year were:

  • Automated Employee Entrance Theme Music in the Office using OpenCV Facial Recognition
  • Programming and Weight Lifting: Using Computer Vision to Estimate a 1 Rep Max

I’ve documented the abstract and the descriptions below, but if you have any thoughts on either topic, PyCon related or not, I’d love to see them in the comments. Also, I guess I’m free to edit the proposals, so if anything is unclear, let me know as well.

Talk 1: Automated Employee Entrance Theme Music in the Office using OpenCV Facial Recognition

This talk would be really easy to do since I’ve already done all the work. You can see a blog post about it here

Description

face_rec2

Facial recognition can be a relatively easy problem to tackle using OpenCV, and in conjunction with a DSLR camera and a Sonos sound system, automated introductions can be created on a per-employee basis for the entire office.

Abstract

When you walk through the entrance doors in any typical office setup, an immediate and instinctive visualization is triggered as your mindset reflexively transforms in preparation for the gold that is to flow from your fingertips and into the keyboard. That song begins to play in your head — that same one that you’re obsessed with right now that you’ve been listening to over and over for the last 3 months that no one knows about except for everyone on Facebook because you forgot to turn your Spotify settings to private. The ground cracks beneath your every step, and buildings explode behind you as you steadily approach your desk. You’ve coded to this song no less than 30 times, so you know when the beat drops, and you time it perfectly to arrive at your seat as your coffee cup turns to dust in your hand while the crowd goes wild. It’s the opening act before you unlock your computer and unleash a torrent of beautiful logic upon your machine.

Screen Shot 2014-09-16 at 11.37.11 AM

It’s what goes through every engineer’s head. But just a little bit of hardware and OpenCV can bring this ideation to life. A set musical theme and a corresponding introduction from professional voice talent should be just another standard employee benefit of a software company and not something obtainable only for the elite few professional wrestlers and UFC fighters.

face_rec1

Talk 2: Programming and Weight Lifting: Using Computer Vision to Estimate a 1 Rep Max

Update: I have a proof of concept for the start of this talk idea here

The work for this project hasn’t been done yet, but I intend to complete it regardless of an accepted PyCon talk. This idea came to me as I was playing with some blocks with my daughters, and then BAM, epiphany. More blog posts will follow.

Description

mvi_8443_Snapshot

For any weight lifter, establishing an exact weight for a given exercise to perform a 1 rep max is usually only accurate to a few percent, and using software to augment one’s training is virtually unheard of. However, using a little bit of computer vision and basic physics, it is possible to determine the exact amount of force exerted upon a weight.

Abstract

The amount of force generated in a real world scenario can generally be calculated using Newton’s Second Law of Motion as long as mass and acceleration in a given direction can be measured. For weight lifters, it would be possible to ascertain an estimated and accurate 1 rep max if the amount of force exerted is known.

In such a setup, mass is generally known, but acceleration isn’t so easy to measure. However, a little bit of computer vision and python (not so widely used among power lifters for some reason) can be used to determine an acceleration vector in pixels per second per second, and pixels can be converted to meters using the relative size of objects in the image (a 45 lb plate on the bar for example).

In this talk we will examine the project itself and the code behind it. From there, we’ll example real-world footage from professional power lifters and bodybuilders to determine the amount of force generated and the associated poundage it represents.