Machine Learning With A Bird Feeder 1 of n

·

2 min read

TLDR; I am no Brandon Rohrer and this is not an End-To-End-ML Course .

Around the time the Covid-19 lock downs started, I went down to the basement with an order from the boss to clean things up. Like Aladdin discovering the magic lamp, I found this beautiful pineapple bird feeder. Ordered some bird feed and hung it out on the deck. Surprisingly, it took a couple of days before the birds started showing up. And oh boy! What a riot once they did. We had great time trying to identify the birds and refilling the feeder as soon as it ran low.

BirdFeeder.png

We took a couple of pictures and tried to identify the birds. This got me thinking. Why not hook up a camera and have it identify the birds? Had a spare Wyze Cam. Unfortunately this camera did not capture pictures with good resolution. In addition, I don't have API level access to the pictures.

The next step was to build my own camera setup, and what better board than a Raspberry Pi? This is one of the most versatile computer boards. I had an older version of Pi Zero without camera access. So bought a new Pi Zero. The next problem I ran into was, the camera cable I had was not compatible. Another round of Amazon shopping.

I now have a programmable camera. In a series of blog posts, I intend to capture my journey of implementing a Bird Identifier system. I am no trained Data Scientist. These posts will be light on math (if any at all), but will capture the essence of building a pipeline using resources from Azure, AWS or GCP as appropriate. The programming language I intend to use is Python (I am a newbie). I'll be collecting my thoughts here on why I decided to do what I did, lessons learned, etc. Each post will focus on one aspect of the problem and the corresponding solution. In addition, I also intend to stream some of the work on Twitch

I hope you will join me on this journey. Stay tuned!