universvm / BacXeption

Deep Learning Template for bacterial image classification in Keras.

Deep Learning for bacterial classification

BacXeption

BacXeption is a Deep Learning template of image segmentation functions and a Convolutional Neural Network (CNN) built on Keras for bacterial image classification. It uses the Xception architecture with pre-trained weights (https://arxiv.org/abs/1610.02357).

Examples

1. Getting Started

This project requires Python 3.6+

1.1 Pre-requisites

Install the prerequisites with PIP

pip install -r requirements.txt

1.2 Running the trained model

1. Place the raw images in data/test_data/
2. Run python main.py

This should output labelled images with a .txt file of the coordinates of each box in the output/$DATE_TIME folder. Example:

2. Training your own model

2.1 Two categories

1. Replace the images in the data/0/ and data/1/ with your images.
2. Run python train.py
3. Move the output/$DATE_TIME/model.json and output/$DATE_TIME/model.h5 in the model/ folder.
4. Follow the instructions in section 1.2

2.1 >Two categories

1. Change NUM_CLASSES in config.py to the number of classes wanted.
2. Add your data in the data/ folder. Each category should have a separate folder name, these must be integers starting from 0 (eg. 0/,1/,2/ for 3 categories)
3. Follow the instructions in section 2.1

3. Contributing

Pull requests and suggestions are always welcome.

4. Additional information

Authors

Leonardo Castorina - universVM

Acknowledgments

Dr. Teuta Pilizota - Proposing the problem and useful discussions.

Dario Miroli – For introducing me to Keras and debugging early versions of BacXeption

François Chollet – Developing Keras and Xception