rishavpramanik / CervicalFuzzyDistanceEnsemble

A fuzzy distance-based ensemble of deep models for cervical cancer detection published in Computer Methods and Programs in Biomedicine, Elsevier

Cervical cancer detection from Pap Smear Images

"A fuzzy distance-based ensemble of deep models for cervical cancer detection" published in Computer Methods and Programs in Biomedicine (June 2022), Elsevier

@article{pramanik2022fuzzy,
title = {A fuzzy distance-based ensemble of deep models for cervical cancer detection},
author={Pramanik, Rishav and Biswas, Momojit and Sen, Shibaprasad and de Souza J{\'u}nior, Luis Antonio and Papa, Jo{\~a}o Paulo and Sarkar, Ram},
journal = {Computer Methods and Programs in Biomedicine},
volume = {219},
pages = {106776},
year = {2022},
issn = {0169-2607},
doi = {10.1016/j.cmpb.2022.106776},
url = {https://www.sciencedirect.com/science/article/pii/S0169260722001626}
}

A fuzzy distance-based ensemble of deep models for cervical cancer detection

Find the original paper here.

Datasets Links

1. SIPaKMeD SCI Pap Smear Images
2. Herlev
3. Mendeley LBC

Instructions to run the code

Required directory structure: (Note: train and test contains subfolders representing classes in the dataset.)

+-- data
|   +-- train
|   |   +--class A
|   |   +--class B
|   |   ...
|   +-- test
|   |   +--class A
|   |   +--class B
|   |   ...
+-- main.py

1. Download the repository and install the required packages:

pip3 install -r requirements.txt

2. The main file is sufficient to run the experiments. Then, run the code using linux terminal as follows:

python3 main.py --data_directory "data"

Available arguments:

• --num_epochs: Number of epochs of training. Default = 70
• --learning_rate: Learning Rate. Default = 0.0001
• --batch_size: Batch Size. Default = 16
• --path: Data Path. Default= './'
• --kfold: K-Fold, to perform K fold cross validation. Default= 5

3. Please don't forget to edit the above parameters before you start