This repository contains a review of the paper "Conditional Random Feilds as Recurrent Neural Networks" that I did in the course Probabilistical Graphical Models. It contains the following files:
- PDF with my written review.
- A Presentation of the previous docuemnt.
- Folder with the code.
The project is intended to be used in Google Colab. For this, either download the folder itself and upload it to Google Drive, or create a new notebook and clone the repository.
Download the model weights from here and place it in the crfasrnn_pytorch directory with the file name crfasrnn_weights.pth.
When imported, open the crfasrnn.ipynb file. This contain all code you will need to run. The only thing that need to be changed are the paths to acces the files, as this may vary. The tested images and their corresponding segmented results will be saved in the folder "samples". In this folder you can place all the images you would like to try.
The code has been extrected form the Github repository of the paper. For more details, acces: https://github.com/sadeepj/crfasrnn_pytorch