Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data rename reproduce to preprocessed Jun 14, 2019
optimaltransport @ 764e8c1 add optimaltransport submodule Jan 24, 2019
.gitmodules
README.md
dataset.py
main.py
models.py add confusion matrix May 16, 2019

README.md

Cell Image Classification

Code for training and testing of a set of statistical machine learning models on the 2D HeLa dataset (https://ome.irp.nia.nih.gov/iicbu2008/hela/index.html).

Dependencies

Usage

usage: main.py [-h] [--dataset DATASET] --space {image,wndchrm,rcdt} --model
               {RF,KNN,SVM,LR,LDA,PLDA,MLP,ShallowCNN,VGG16,InceptionV3}
               [--transfer-learning] [--SVM-kernel {rbf,linear}] [--preprocessed]

P1 Cell Image Classification

optional arguments:
  -h, --help            show this help message and exit
  --dataset DATASET
  --space {image,wndchrm,rcdt}
  --model {RF,KNN,SVM,LR,LDA,PLDA,MLP,ShallowCNN,VGG16,InceptionV3}
  --transfer-learning   neural network use pretrained weights instead of
                        training from scratch
  --SVM-kernel {rbf,linear}
  --preprocessed           reproduce the results on Hela dataset reported in the
                        paper

Examples

  • Train A logistic regression model on image space: python main.py --space image --model LR

  • Train A logistic regression model on WND-CHARM feature space: python main.py --space wndchrm --model LR

  • Train InceptionV3 on image space: python main.py --space image --model InceptionV3

  • Train InceptionV3 on image space by fine-tuning a pre-trained model (transfer learning): python main.py --space image --model InceptionV3 --transfer-learning

Reproduce Hela Results

We provide the data used for producing the Hela results as reported in the paper. The preprocessed data is located in data/hela_preprocessed. To reproduce the results, add the --preprocessed flag to commands (this will load the propressed data). For example, to reproduce PLDA classification on wndchrm features result, run python main.py --dataset hela --space wndchrm --model PLDA --preprocessed.

You can’t perform that action at this time.