CSCI 8360 Data Science Practicum : Project NeuroFinder

Goal

To develop a segmentation pipeline that identifies the regions of as many neurons in the video as possible, as accurately as possible.

Getting Started

These instructions describe the prerequisites and steps to get the project up and running.

Setup

This project can be easily set up on the Google Cloud Platform, using a 'Compute Engine VM Instance' or a 'Deep Learning VM Instance' with/without a GPU. You will need to have Google Cloud SDK installed in your local machine to be able to set the project up.

After downloading/cloning this repository to your local machine, the user will need to open the Google Cloud SDK Shell. Once it opens, the user can copy the contents of this repository to the Deep Learning VM instance using the command:

gcloud compute scp --recurse /complete/link/to/repository/* <user>@<instance_name-vm>:/home/<user>/

Furthermore, to setup a VM instance with all the prerequisite packages used in the project, do the following:

1. Install Conda using the conda_install.sh file in the 'scripts' directory using the command: $ bash scripts/conda_install.sh

2. Create a conda environment from the included environment.yml file using the following command:

   $ conda env create -f environment.yml

3. Activate the environment

   $ conda activate hopper

Usage

To run the code and generate output prediction masks in the /results directory, the user can navigate to the folder containing the file 'team-hopper.py', and run it using the command: $ python team-hopper.py --options. The user can get a description of the options by using the command: $ python team-hopper.py -help.

Output

Upon running the command in the ‘Usage’ section, the dataset will be downloaded from the Google Storage bucket link carrying the Neuron dataset, and the output submission.json files will be generated in the /results directory. Here's a preview of the submission.json file:

 [                                                   
     {  "dataset": "00.01.test",                      
        "regions":
           [
             {"coordinates": [ [0,0], [0,1] ]},
             {"coordinates": [ [10,12], [14,17] ]},
           ]                                           
     }                                                    
 ]

Ethical Implications

These models should only be used to better science and medicine - it should not be used to develop any biological weapons or any other sorts of weapons which can cause harm to the society.

Contributors

• See Contributors file for more details.

License

This project is licensed under the MIT License. See LICENSE for more details.