What it does
The app downloads all images from an Instagram user's posts. It then identifies the most common face and crops it to only show that person's face. The app then uses a machine learning model to predict the likelihood of each of the cropped images being an adult or a child. It then outputs the prediction and the likelihood of it being accurate.
How I built it
I used Instaloader to download all the posts from a user. I used OpenCV to identify the most common face and to crop it. I then used PyTorch to train and load the model. I used tkinter to build a very rudimentary interface.
Challenges I ran into
I was working with a limited dataset, as I could not find a more comprehensive one. I also struggled with technical issues and incompatibility with many of my libraries and had to switch the modules and tools that I was using entirely to get around it.
Accomplishments that I'm proud of
I learned how to use PyTorch to create an ML model. I used OpenCV for the first time and I want to explore its potential in greater depth in my future projects.
What I learned
How to use PyTorch and OpenCV
What's next for InstaVerify
Improving the GUI
Improving speed
Automating functionality
Improving error-handling
Log in or sign up for Devpost to join the conversation.