Large-Scale Multi-Class Image-Based Cell Classification with Deep Learning

Tensorflow implementation of "Large-scale Multi-class Image-based Cell Classification with Deep Learning" by Nan M., Edmund Y. L., Tsia K. M., Hayden K-H. So. [Project]

Project page

https://monaen.github.io/Meng2018Largescale

Environment

• Ubuntu 16.04
• Python 2.7
• Note: for Windows Users you need to download the dataset and unzip to the "data" folder.

Code

We provide the training and test code at: https://github.com/monaen/Meng2018Largescale/tree/code

Dataset

We provide three different datasets, and all the datasets could be downloaded at https://github.com/monaen/Meng2018Largescale/tree/data

• download the small data

git clone https://github.com/monaen/Meng2018Largescale.git --branch code --single-branch
cd Meng2018Largescale/data
./download_small_Train_Test_Valid.sh

• download the large data

./download_large_Train_Test_Valid.sh

• download the augmented data

./download_augmented_Train_Test_Valid.sh

• run the demo

python demo.py

Details

The entire code project contains many recent deep learning models for cell classification. We implement these models according to their papers and integrate them into our project.

• cellnet: our work
• gaohep2net: Gao et al.("HEp-2 cell image classification with deep convolutional neural networks")
• liuhep2net: Liu et al.("HEp-2 cell classification based on a Deep Autoencoding-Classification convolutional neural network")
• oeinet: Oei et al.("Convolutional neural network for cell classification using microscope images of intracellular actin networks.")

If you want to use our data or code please cite our paper properly.

Reference

Data

@data{h2qw97-18,
  doi = {10.21227/H2QW97},
  url = {http://dx.doi.org/10.21227/H2QW97},
  author = {Nan Meng; Edmund Lam; Tsia; Kevin Kin Man; So; Hayden Kwok-Hay },
  publisher = {IEEE Dataport},
  title = {Human somatic label-free bright-field cell images},
  year = {2018}
}

Paper

@article{Meng2018Largescale,
  title = {Large-scale Multi-class Image-based Cell Classification with Deep Learning},
  author = {Meng, Nan and Lam, Edmund and Tsia, Kevin Kin Man and So, Hayden Kwok-Hay},
  journal = {IEEE journal of biomedical and health informatics},
  year = {2018},
  publisher = {IEEE}
}

TODO

• Update the proposed model -- CellNet
• Upload the small/large dataset
• Upload the augmented dataset
• Upload the well-trained weights for CellNet
• Add the Hep2 model from Gao et al.("HEp-2 cell image classification with deep convolutional neural networks")
• Upload the well-trained weights for Gao et al.'s model.
• Add the Hep2 model from Liu et al.("HEp-2 cell classification based on a Deep Autoencoding-Classification convolutional neural network")
• Upload the well-trained weights for Liu et al.'s model.
• Add the model from Oei et al.("Convolutional neural network for cell classification using microscope images of intracellular actin networks.")
• Upload the well-trained weights for Oei et al.'s model.