/NVIDIA-image-seg

Primary LanguageJupyter Notebook

Getting Started with Image Segmentation

This is a simple IPython Notebook created for the Nvidia Deep Learning Institute course on image segmentation.

Tasks

  1. Creating a simple NN for image segmentation (data preprocessing)
  2. Using CNN
  3. Implementing Dice Metric and parameter search

Files:

image segmentation.ipynb:

  • Using simple NN for image segmentation (inefficient)
  • Using CNN with multiple convolutions and one convolutional transpose layer

UNET+Dice.ipynb:

  • Using the previous CNN with Dice coefficient which provides a better comparison of detected area with the actual area
  • Implementing a U-Net architecture with encoder-decoder layers