Okay, so let me tell you about my WWE Ava project. It was a wild ride, but I learned a ton.
It all started when I was looking for a fun side project to get my hands dirty with some new tech. I stumbled upon some tutorials about image recognition, and I thought, “Hey, why not try to build something that can identify WWE wrestlers?” And since Ava Raine (Simone Johnson) is super interesting, I decided to focus on her.
First things first, I needed data. And I mean a LOT of data. So, I started scraping images of Ava from all over the internet. Google Images, WWE’s website, social media… you name it. I spent a good couple of evenings just downloading and organizing images into different folders. It was super tedious, but absolutely necessary.
Once I had a decent dataset, I started playing around with some pre-trained models. I figured I wouldn’t reinvent the wheel. I used TensorFlow and Keras because I was most familiar with them. I tried a few different architectures, like VGG16 and ResNet50. I quickly realized that the accuracy wasn’t great out of the box.
Then I did some serious data augmentation. I rotated, cropped, flipped, and zoomed in on the images. Basically, I tried to create as much variation as possible from the existing dataset. This made a HUGE difference. The model started to learn the key features of Ava’s face better.
The training process was a pain, I won’t lie. My old laptop was chugging, and it took forever to run each epoch. But eventually, after lots of tweaking and experimenting with hyperparameters, I managed to get the accuracy up to around 85%. Not perfect, but good enough for a personal project.
To test it, I threw in some random images of other female wrestlers and celebrities. It was surprisingly accurate! It correctly identified Ava most of the time, and it didn’t falsely identify her too often. I even tried some images with bad lighting and different angles, and it still did pretty well.
The next step was to build a simple web app around it. I used Flask for the backend and a basic HTML/CSS frontend. Users could upload an image, and the app would tell them if it thought it was Ava or not. It was super basic, but it worked!
This project taught me a lot about image recognition, data augmentation, and even a bit about web development. It was a lot of work, but definitely worth it. Plus, it’s just cool to have a little AI that can recognize one of my favorite WWE personalities.
Of course, this is just the beginning. There’s so much more I could do with it. Maybe I’ll try to train it on more wrestlers, or even try to detect different poses or expressions. Who knows? The possibilities are endless!
