Contamination_DenseNet

Contamination classification for explants

This repository contains the pipeline for classifying explants into categories - contaminated, non-contaminated, missing

This is Keras implementation of DenseNet. The code for densenet used in this repository is obtained from here.

To know more about how DenseNet works, please refer to the original paper

Densely Connected Convolutional Networks
Gao Huang, Zhuang Liu, Kilian Q. Weinberger, Laurens van der Maaten
arXiv:1608.06993

Setup environment.

1. Create a conda/python virtual environment.

2. Install the dependencies from requirements.txt using pip -

    conda create --name <envname> --file requirements.txt

To manually install dependencies, follow the below steps -

1. Create a conda create environment:

conda create -n envname  python=3.6
conda activate test1

2. Install Keras and Tensorflow-gpu using Conda

conda install -c conda-forge keras
conda config --set restore_free_channel true
conda install tensorflow-gpu=1.13

4. Pip install libraries

pip3 install opencv-python Pillow matplotlib easydict argparse tqdm

Using Conda

conda install -c conda-forge opencv matplotlib easydict argparse tqdm
conda install -c anaconda pillow

To run predictions on the dataset, please follow the below steps.

1. Create a directory for the input rgb images, and place the dataset into this directory.

   Please note that currently the code only supports grid type 12 (3 X 4).

2. Please place the downloaded model in the main working directory.

3. Create a CSV which contains the list of input RGB images. Check test.csv in the folder.

4. To run the script, use the command -

Please update the output directory path in the config.py file (OUTPUT)

    KERAS_BACKEND=tensorflow python inference.py --img-list test.csv 

Or, Include img-list, crop_dims and output_file paths as arguments to the script.

    KERAS_BACKEND=tensorflow python inference.py --img-list test.csv --output_file output_file_name --crop_dims "(260, 600, 1700, 1710)"  

To turn on debugging, use the debug flag and set it to True. (--debug True)

Format of the output CSV -

image_name,1,2,3,4,5,6,7,8,9,10,11,12
GWZ7_I2.0_F1.9_L80_194153_8_1_3_rgb.jpg,NC,NC,NC,NC,NC,NC,M,NC,NC,NC,NC,NC
GWZ7_I2.0_F1.9_L80_194927_1_2_5_rgb.jpg,C,NC,NC,NC,NC,NC,NC,NC,NC,NC,C,NC
GWZ7_I2.0_F1.9_L80_194429_11_2_2_rgb.jpg,NC,NC,NC,NC,NC,C,NC,NC,NC,NC,NC,NC
GWZ8_I2.0_F1.9_L80_195913_4_0_6_rgb.jpg,C,NC,NC,C,C,C,C,NC,C,C,C,NC

Requirements

• Keras 2.0.5