RealTimeBird is a tool designed to analyze video footage and pinpoint the moments when birds – or even squirrels – make their appearances. And the best part? It creates clips of these special moments automatically.

“Why can’t I eat the bird food if I was labeled as a bird?”

So, here's the story: I've recently found myself really into bird watching. I decided to set up a feeding station at my place to see what kinds of feathered friends I could attract. To satisfy my curiosity about what birds were dropping by and when, I set up a camera to livestream and record all the action. Everything was saved onto my personal storage server.

Now, the camera was capturing everything, all day, every day. But as you might expect, our birds were only making guest appearances for tiny fractions of that time. I found myself tediously scrolling through hours of footage to find those precious moments when the birds came to visit, then manually trimming these instances into clips. Talk about a time-consuming task! Plus, with all that video, my storage server was quickly hitting its limits. I needed to fix this.

A little bit of research led me to the YOLOv5 image recognition model and the OpenCV python library. I realized that if I fed individual video frames into this model, I could calculate exactly when the birds were popping in for their meals and drinks. Then, with the command-line tool FFmpeg, I could generate individual clips of each bird visit. Problem solved!

And that's the story behind RealTimeBird. The name might throw you off a bit because I initially wanted to run the program in real time, especially as the model could handle it. The idea was to only trigger recording when the birds were visiting. But I quickly realized this approach came with extra costs, like needing a remote shutter cable and possibly even an Arduino controller. So, I put that thought on hold and decided to start with this post-processing version instead.

As for what's coming up, I've got some exciting ideas for future improvements. In addition to detecting bird visits and creating clips, I'm planning to dive deeper and identify the species of birds that came by. But I'm not stopping there! I'm thinking about doing some fun stuff like gathering statistics on the variety of bird visitors and seeing how different types of weather or food affect each species. So, keep your eyes peeled for the updates - there's more to come!