Setup Instructions

Model Weights

1. Download SEM TEM Other Classifier weights and place it in classifier/SEM_TEM_Other_weights.
2. Download Particulate Non-Particulate Classifier weights and place it in classifier/Particulate_nonParticulate_weights.
3. Download Figure separation weights and place it in figure-separator/data.
4. Download SRCNN weights and place it in label_scale_bar_detector/OCR/SRCNN-pytorch/weights/.
5. Download Darknet weights and place it in label_scale_bar_detector/localizer/darknet/backup.
6. Download Mask RCNN weights and place it in particle_segmentation/Mask_RCNN/logs/tem.

Installation

If you would like to run the entire pipeline, Run conda env create -f environment/environment.yml.

If you would only like to download the datasets, Run conda env create -f environment/environment_dataset.yml.
Note: These installations have been tested only on a Linux system.

Datasets

Downloading JSON files

1. The json file with all extracted size/shape information corresponding to the 4361 literature-mined images can be downloaded from Full_dataset.
2. The json file with segmentation annotations corresponding to 131 images used as training data for the Mask-RCNN can be downloaded from Training_dataset.

Place both files at the root of the repository.

Downloading images

1. To download the full literature-mined dataset of 4365 images, run python fetch_urls_full_dataset.py.
2. To download the annotated dataset of 131 images that was used to train the segmentation model, run python fetch_urls_training_dataset.py.

Running the pipeline

Run python test_pipeline_single.py.

The following is an illustration of the steps involved in the pipeline.

Acknowledgements

1. https://github.com/AlexeyAB/darknet
2. https://github.com/apple2373/figure-separator
3. https://github.com/yjn870/SRCNN-pytorch
4. https://github.com/matterport/Mask_RCNN

Citation

If you use this code, please cite the following manuscript:

@misc{subramanian2021dataset,
      title={Dataset of gold nanoparticle sizes and morphologies extracted from literature-mined microscopy images}, 
      author={Akshay Subramanian and Kevin Cruse and Amalie Trewartha and Xingzhi Wang and Paul Alivisatos and Gerbrand Ceder},
      year={2021},
      eprint={2112.01689},
      archivePrefix={arXiv},
      primaryClass={cond-mat.mtrl-sci}
}