[WIP] [Detailed Readme to be added]
[ToDo]
- Update the Readme -> Provide detailed instructions on how to run and where to make changes -> List dependencies
- Use a different loss than MSELoss
- Remove unnecessary code/comments and add more helpful comments
- Implement code to train and evaluate on part of dataset (in between epochs) and make it a parameter
- Try more transforms
- Make displaying images a parameter
- Nice graphs -> Possibly use Visdom or Tensorboard
Convolutional Encoder Decoder network based on the SegNet architecture for unsupervised feature learning
This SegNet based Convolutional Encoder-Decoder network can be used for unsupervised feature learning for particular datasets. The PyTorch Dataset class for the MICCAI dataset has been provided. You can make the changes in that to adapt to your dataset.
A sample training script is also provided.