pneumonia_detector-beyond

• 

  Web app

Inspiration

Every day nearly 2,200 children die from Pneumonia under the age of 5 (Over 15% of all deaths under the age of 5). Pneumonia is a curable and preventable disease. If diagnosed in time, these deaths can be prevented.

We built this project to help solve this problem.

What it does

Our project detects pneumonia in X-rays using computer vision and machine learning with 92% accuracy.

How we built it

We trained a densenet121 model with 121 trainable layers with pytorch using pneumonia dataset from kaggle.

Challenges we ran into

Selection of the model was key to getting high accuracy. Tuning hyperparameters and training a model in a day was challenging.

Accomplishments that we're proud of

We have successfully trained a classifier for pneumonia with 92% accuracy and deployed a web app in less than 24.

What we learned

We learned multiple things about neural networks and training neural networks as well as deploying them.

What's next for penumonia_detector-beyond

We hope our detector will encourage more utilization of machine learning in providing better healthcare and saving lives.

Built With

• computer-vision
• machine-learning
• pytorch