TensorFlow is an end-to-end open source platform for machine learning. It has a comprehensive, flexible ecosystem of tools, libraries, and community resources that lets researchers push the state-of-the-art in ML and developers easily build and deploy ML-powered applications.
TensorFlow was originally developed by researchers and engineers working on the Google Brain team within Google's Machine Intelligence Research organization to conduct machine learning and deep neural networks research. The system is general enough to be applicable in a wide variety of other domains, as well.
TensorFlow provides stable Python and C++ APIs, as well as non-guaranteed backward compatible API for other languages.
We will be using Tensorflow 2.1.0 with Python3 for detecting Covid-19 Pneumonia signs from CT Scan Images by a CNN(Convolutional Neural Network) Model. The model have a uniform dataset of 764 Images of CT Scan which consist 349 Images of Covid-19 Pneumonia affected patients and remaining shows normal patient scans.
This project should be used for research purposes only. The purpose of the project is to show the potential of Artificial Intelligence for medical support systems such as diagnosis systems. Although the program is fairly accurate and shows good results both on paper and in real world testing, it is not meant to be an alternative to professional medical diagnosis. The developer has experience in using Artificial Intelligence for detecting COVID-19, but is not a doctor, medical or COVID-19 expert. Please use this system responsibly.
Please follow the Installation Guide to install COVID-19 AI Classification Tensorflow.
The Project can be used for following purposes:
- Training a CNN model on the given Image Dataset
- Unzip the provided CT Scan Image dataset provided in Classes directory or you can also generate or import any other CT Scan dataset in this directory with same directory structure as in unzipped dataset folder.
- Use dataset_preparation.py to resize all the images into 150x150 Pixels, then converting the PNG format dataset and their labels(covid19 or normal) into Pickle as the Model will take them in form of pickle dataset.
- The above program will generate two pickle files(X.pickle, Y.pickle) in same Classes directory, we have also provided generated pickle files for same dataset, so that you can use these pickle files directly in generating CNN Model by using init.py.
- The program(init.py) will generate a saved model file and a tensorboard log file which shows the performance of model loss and accuracy in scalar graph.
Note: Use rename_image_and_filetype_conversion.py for making all images in same format(PNG) and indexed as it is possible that while using a new or any open source CT Scan Image dataset in Step 1, all may not be in same Image format and Indexed properly.
- Predicting CT Scan Image
- After training the model on the dataset, use predict_ct_scan.py to test on any lung ct scan image, the program will first resize the image to 150 x 150 pixel size, then loads the image into model and print the type of scan(Covid-19 or Normal Scan). It should be noted that the prediction of model is based on these parameters: Training Accuracy, Training Loss, Validation Accuracy and Validation Loss for the given dataset and the trained model architecture.
- Using Tensorboard for Optimization, Analysing and Selection of CNN Architectures
- Add/Change the parameters for conv_layers, conv_sizes and dense_layers of init.py and run it in shell(Comment the last line of program if you don't want to save the trained model) in Classes Directory.
- A new folder named: "logs" will be genearated in the same Directory. Now run shell in same Classes directory again and execute the following command:
tensorboard --logdir logs
Open your browser and navigate to shown web address in shell. In our case we gave many values to the list of parameters and the tensorboard tab was recorded with the following results:
Open source Artificial Intelligence (CNNs & GANs) for COVID-19 Pneumonia detection/early detection.
Detects Covid-19 Pneumonia signs from CT Scan Images by a CNN Model. The model have a uniform dataset of 764 Images of CT Scan which consist 349 Images of Covid-19 Pneumonia affected patients and remaining shows normal patient scans.
Here are some CT-Scans of lungs:
COVID19 Patient Scan:
Normal Patient Scan:
The dataset was taken from the following source. You can go through our dataset where we have indexed all the images and converted them into same format(PNG). We have used tensorflow library for training a binary classification model of CT Scans using Convolutional Neural Network. The graph of model is as follows:
We have also uploaded 2 trained models on above graphs:
- Model 1: Validation Accuracy - 80.82%, Validation Loss - 50.10%
- Model 2: Validation Accuracy - 78.00%, Validation Loss - 47.28%
As the dataset of CT Scans of COVID19 patients is limited, the model seems to overfit, you can follow the instructions for contributing on CONTRIBUTING.md
The Peter Moss COVID-19 AI Research Project encourages and welcomes code contributions, bug fixes and enhancements from the Github.
Please read the CONTRIBUTING document for a full guide to forking our repositories and submitting your pull requests. You will also find information about our code of conduct on this page.
- Aniruddh Sharma - Asociacion De Investigation En Inteligencia Artificial Para La Leucemia Peter Moss R&D, Ahmedabad, India
We use SemVer for versioning. For the versions available, see Releases.
This project is licensed under the MIT License - see the LICENSE file for details.
We use the repo issues to track bugs and general requests related to using this project. See CONTRIBUTING for more info on how to submit bugs, feature requests and proposals.
