50 Shades of Grey

• 

  Pneumonia Chest X-Ray

• 

  Normal Chest X-Ray

• 

  Data Visualization in Notebook

• 

  Training the Machine Learning Model

• 

  Using DCP Parallelization for Data Pipelines

• 

  Model Validation

Contact

Discord: itssujee#9904

Team Number: 65

Inspiration

During the COVID-19 pandemic hospitals have been over-working their staff and as a result quality of care has taken a hit. One of the common ailments associated with contracting COVID-19 is pneumonia, which is a deadly combination.

What it does

The 50 Shades of Grey Project uses a Machine Learning Model to determine whether or not an X-ray shows symptoms of pneumonia. Much of the backend processing is done through DCP, which allows for quick distributed cloud processing.

How I built it

The dataset was hosted on the Google Cloud Platform in the GCS bucket. I made the dataset public to allow for easy access in my notebook. I found a couple articles, research papers, and official Tensorflow documentation to create the Machine Learning Model. I used the DCP API to distribute the work required to process the data for the machine learning model to digest it.

Challenges I ran into

The DCP API does not support a lot of features that data scientists would find useful. Furthermore the platform is in a extremely beta form and the lack of documentation made the discovery process more difficult.

Accomplishments that I'm proud of

I'm proud that my model achieved a 98% recall rate, and a 70% accuracy on the testing dataset.

What I learned

I learned how to data engineer using parallelization, as well as I figured out how to build, train, and export a machine learning model in Tensorflow.

What's next for 50 Shades of Grey

Developing better support for parallelization for data scientists in conjunction with the DCP Platform.

Built With

• colab
• dcp
• google-cloud
• keras
• matplotlib
• node.js
• numpy
• pandas
• sklearn
• tensorflow