Okay, so today I wanna talk about something I messed around with recently: gionna daddio. Now, before you get the wrong idea, it’s not what you think! I was just experimenting with image recognition and thought it would be a fun test case.

It all started when I was trying to build a simple object detection model. I’d been using the usual datasets, you know, cats, dogs, cars – the boring stuff. I wanted something a bit more… unique. So, I thought, “Why not try recognizing faces? And why not make it a specific face?”
First, I needed data. I spent a good chunk of an afternoon scraping images. I mean, seriously, just Googled and Binged my way through tons of pics. It was tedious, I won’t lie. And honestly, sorting through them all took even longer.
Next up was the labeling. This part was a pain in the butt. I used a free tool I found online to manually draw bounding boxes around her face in each image. It was super repetitive, and my eyes started to cross after a while. Seriously considered giving up at this point. But, I persevered!
Then came the fun part – actually building the model! I went with a pre-trained model and fine-tuned it on my dataset. I’m no expert in machine learning, so I stuck with something relatively simple. There was some tinkering involved, adjusting the learning rate and other hyperparameters, until I got something that seemed to work decently.
Finally, the moment of truth: testing it out. I fed it some new images it hadn’t seen before. At first, the results were… mixed. Sometimes it nailed it, sometimes it completely missed. There were even a few false positives where it thought other people were her which was kind of funny.

So, I went back and did a bit more tweaking, added a few more images to the dataset, and retrained the model. After a few iterations, it got a lot better. Not perfect, mind you, but good enough for a fun little project.
What did I learn? Well, for one, data collection and labeling is a HUGE time sink. Seriously underestimated that. Also, pre-trained models are your friend! Saved me a ton of time and effort. And finally, even a simple project can be a great way to learn about machine learning concepts.
Would I do it again? Maybe. But next time, I’ll probably pick something a bit less… controversial. Lesson learned!