npsam 1.2

NP-SAM is an easy-to-use segmentation and analysis tool for nanoparticles and more.

NP-SAM

Introduction

In this project we propose an easily implementable workflow for a fast, accurate and seamless experience of segmentation of nanoparticles.

The project's experience could be significantly enhanced with the presence of a CUDA-compatible device; alternatively, Google Colab can be utilized if such a device is not accessible. For a quick access to the program and a CUDA-GPU try our Google Colab notebook.

installation

Clone the repo with git and enter the newly created np-sam directory. Just copy the code and paste in the terminal for macOS or in Anaconda Prompt for Windows:

git clone https://oauth2:[e-mail]/disorder/np-sam.git
cd np-sam

Alternativly, to circumvent Git, download the files manually from the repo and place in a directory called np-sam and enter the directory with cd np-sam.

Create a new conda environment called NPSAM and activate it.

conda create -n NPSAM python=3.10
conda activate NPSAM

Install PyTorch.

Now install the remaining required packages, and make a static link to the ipykernel

pip install -r requirements.txt
python -m ipykernel install --user --name NPSAM --display-name NPSAM

Finally, download the weights for SAM from their GitHub page or with the following command:

curl -O https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth

Get started

In the working directory and NPSAM environment execute jupyter lab in the terminal. This will launch jupyterlab.

Citation

@article{NPSAM,
   author = {Larsen, Rasmus and Villadsen, Torben L. and Mathiesen, Jette K. and Jensen, Kirsten M. Ø and Bøjesen, Espen D.},
   title = {NP-SAM: Implementing the Segment Anything Model for Easy Nanoparticle Segmentation in Electron Microscopy Images},
   journal = {ChemRxiv},
   DOI = {10.26434/chemrxiv-2023-k73qz-v2},
   year = {2023},
   type = {Journal Article}
}

Acknowledgment

This repo benefits from Meta's Segment Anything and FastSAM. Thanks for their great work.