Tensorflow for Computer Vision Project

Leveraging TensorFlow to create bounding boxes and classes on CT scans of the lungs.

Data information

• Kaggle, 60GB, 200k images with bounding boxes and classes.

To get started

1. Download the dataset from Kaggle
2. Use the data_prep.py script to create TF records
3. Use the model.ipynb file to create and fit the model

TODO

• Allow the user to start from a previous checkpoint, load save model file.
• Convert all notebooks to python scripts, create classes, and allow the entire process to be kicked off from scipts.
• Define a Dockerfile
• Add functionality to let the user choose their own metrics and loss function
• Show the results from my training job, and provide an analysis
• Test out batch normalization vs instance based normalization
• Write functional tests for this project