th3rring/comp540_final repository overview

⁠COMP 540 Final Term Project

⁠Rahul Popat, Thomas Herring

Kaggle Link⁠

Codelab Link⁠

Project code in project.ipynb Download data and add it to the project locally in a directory called Agriculture-Vision

⁠Docker

In order to run the notebook containers, run either run_gpu_container.sh or run_cpu_container.sh from the scripts directory depending on whether or not you want GPU acceleration. In order to run either, Docker must be installed⁠ and for the GPU containers, nvidia-docker2⁠ must be installed.

All the images are stored in the image directory. When running a container, Docker should try and pull the images from DockerHub⁠ but if not, manually pull the images to your local machine.

You can view all running Docker containers by typing docker ps in a terminal on your machine. To stop a running container, use the appropriate stopping script in the scripts directory.

⁠Accessing Notebooks

Once an image is run, a Jupyter Notebook session will be started on localhost:8888 and a Jupyter Lab session on localhost:8889. The ports for these sessions are set in the Dockerfiles for each associated container. To access these notebooks, visit these URLs in your web browser or click the links below:

Local Jupyter Notebook⁠

Local Jupyter Lab⁠

⁠Running Notebooks

Our most recent implementation of the segmentation model lives in the unet_model.ipynb file. The first few cells define helper functions while the later ones actually train and test our models. Choose a model and run the associated cells!

⁠Online resources

keras Unet official implementation⁠

Unet blog post and explanation⁠