AngelEye AI

Home
Process
Quiz
Xray Model
About Us

Inspiration

Due to the rise of artificial intelligence hype and our team's interest in a shared Machine Learning class last semester, we decided to tackle the challenge with Computer Vision/Image Classification.

What it does

We trained two models that are capable of diagnosing illnesses. One model is a binary classification of Brain Tumors from MRIs (100% accuracy). The other classifies X-ray images into Normal vs Covid vs Pneumonia (97% accuracy). We applied CAM visualization to superimpose a heatmap on each of these.

We then deploy the X-ray model on our front end, where we have a diagnosis page, a quiz section where you play vs the AI, and an Image search section where you upload an x-ray image, get a diagnosis + heatmap, and the application returns relevant case studies and information based on the diagnosis.

How we built it

The models were trained using Transfer Learning with vgg16 as the base model, and some new appended trainable layers.

We built the front-end in Angular.js and a Back-end in Python. We handled API functions with FastAPI.

Challenges we ran into

At first, we were using flask for our API but we ran into a lot of technical issues, eventually forcing us to move to FastAPI.

Since we were on limited time, we couldn't risk trying to train bigger, more ambitious models that would fully demonstrate our Image Search Engine capabilities.

Styling the front end exactly how we wanted was quite time-consuming and we weren't able to complete our vision for the look.

Accomplishments that we're proud of

We were initially very worried about the model training, but we're happy with how accurate our models turned out.

Additionally, the amount of productivity we were able to muster came as a surprise to us, and midway through the project we were able to go bigger with our ideas due to how much progress we were making.

What we learned

We learned to think in industry context and to come up with problem-specific ideas.

We learned technologies and learning techniques that helped us deploy and train our models.

What's next for AngelEye AI

We believe our image search shows a lot of promise and could be a really useful tool for med students and researchers as it expedites the process of finding and analyzing information.

We hope to build a larger database and knowledge in our models, so more diseases and problems can be diagnosable with just a snap of a camera.