Skip to content

Thehunk1206/Covid-19-covidcnn

Repository files navigation

🛑 Note: This model is still not reliable to use. It is just a research for now.

To read Motivation and Goal behind this research go to this Github Repository.

Dataset

dataset
├── test
│   ├── covid19 [32 entries]
│   ├── normal [25 entries]
│   └── pneumonia [20 entries]
├── train
│   ├── covid19 [219 entries]
│   ├── normal [210 entries]
│   └── pneumonia [210 entries]
└── validate
    ├── covid19 [32 entries]
    ├── normal [20 entries]
    └── pneumonia [20 entries]

Soon more data will be added

Model Design

covidcnn161

It is a custom model designed using Residual Blocks. Learn more about Residual block

model_161

  • New model based on Densenet moudule comming soon.

Model Result

Model Accuracy Model Loss

  • This model was trained on very small sets of images, therfore the model is not generalized enough to use.

Usage

$python3 detect.py --help
usage: detect.py [-h] -m MODEL -i IMAGE

optional arguments:
  -h, --help            show this help message and exit
  -m MODEL, --model MODEL
                        path to model
  -i IMAGE, --image IMAGE
                        path to input image

Example

python3 detect.py -m model/trained_model/covid-19_large.h5 -i model/samples/normal.jpeg 

Deploying model

After the model was trained it was compressed into tflite model. After which it was deployed on flutter application.

ScreenShots of Flutter application

ss ss ss

model has been deplyed on a WebApp webApp ss

Dataset source

This repository will be updating as the dataset will grow more and more. Any contribution would be appreciated.

more detailed documentation comming soon

Citation

Paper availabe here

@article{cohen2020covid,
  title={COVID-19 image data collection},
  author={Joseph Paul Cohen and Paul Morrison and Lan Dao},
  journal={arXiv 2003.11597},
  url={https://github.com/ieee8023/covid-chestxray-dataset},
  year={2020}
}