Inspiration
The COVID-19 track caught my eye when I first came to the hack3 website, and I've been spending most of my time at home due to quarantine. I've had some experience with image classification in the past, so I thought, why not?
What it does
The model I've implemented is meant to categorize provided images as either positive or negative for patients.
How I built it
I built it using a custom PyTorch model with a cross-entropy loss function and SGD optimizer.
Challenges I ran into
Tuning hyperparameters to produce the best validation loss, as well as combining datasets.
Accomplishments that I'm proud of
The 44% test accuracy the model attained.
What I learned
That hackathons are little bit nerve-wracking.
What's next for COVID-19 Detector from X-Rays
Fixing the image classification and getting a larger dataset.
Log in or sign up for Devpost to join the conversation.