Covid-19-Dectector

GUI_covid19_detector allows the user to import the image from local machine and with the saved model(already trained model) it performs testing and shows the results either positive or negative.

System Preference

• Dataset Used: covid-19 CT Xray images. • IDE: Jupyter notebook • GPU : AMD Radeon pro 5000M (4 GB GDDR6 ) • CPU: Intel i9 2.3Ghz processor with 16gbram • OS: MAC Os’ BigSur • DCCN Model Tested : ResNet50 and InceptionV3 • Time taken by GPU and CPU for running 10 epoch per model = 10 -15 mins

Proposed Algorithm

1. Import the Resnet50 and inceptionv3 from tensorflow.keras
2. Import the covid-19 dataset
3. Reshape the input image with respect to DCNN model , here 244 ,224 because Resnet and inception will accept 224 x 224 image size .
4. Convert to array and Normalize to interval of [0,1] both positive and negative datasets.
5. Split into training(80%) and testing sets(20%) for both types of images and combine both positive and negative together for both train and test.
6. Build the models: ResNet50 and InceptionV3
7. Start training and calculating accuracy with different optimizer
8. Choose the best model with ‘accuracy’ as performance metrics
9. Save the best model in local machine which will be useful for implementing GUI application .
10. GUI_covid19_detector allows the user to import the image from local machine and with the saved model(already trained model) it performs testing and shows the results either positive or negative.

Trained with 2 different ANN Models with different Optimizers**

1. ResNet50 2)InceptionV3 Model

Image Ref: https://towardsdatascience.com/understanding-and-coding-a-resnet-in-keras-446d7ff84d33

Overall Accuracy with Different Models:

Platform Independent GUI for COVID-19 Detector

Youtube Live Demo

https://youtu.be/eM5_oUZO1Sc