COVID-Detection-with-Ensemble

Setting environment :

Link to the dataset: Download the dataset from this link "https://drive.google.com/drive/folders/1j95dLLZadzyQ-MvOHq_W4Qs_9FpXEAE4". Structure of Dataset:

• Covidx-data
• test (images for testing)
• train (images for training)
• test.txt (annotations and labels for the training data)
• train.txt (annotations and labels for the testing data)

Clone this repository: git clone "https://github.com/pearljain7/COVID-Detection-with-Ensemble.git" Use Kaggle or Google Colab to run the python notebooks on the given dataset. On Kaggle/Google Colab:

• Create a new project.
• Upload the dataset given.
• Once the dataset is uploaded, you will see your input files in the following directory: ../input/covidx-cxr2/
• The directory will contain data in the same format as the drive link.
• Upload the chosen notebook for either of the five models.
• For the ensemble models, also upload the given .txt files (which are prediction arrays from models) to the ../predarrays folder: resnetarr.txt vgg3.txt xception-pred.txt inceptionarr.txt efficientNet-pred.txt Run the notebook titled covid-x-ray-ensemble-and-examples.ipynb.

Find best params using Grid Search:

• Run the chosen model from grid search to gauge the optimal learning rate and Optimiser.
• The grid search runs 10 epochs and iterates over each combination of parameters in the grid.
• If we test 4 learning rates with 3 optimizers, we end up with a total of 12 combinations.
• The entire process might run for 2-4 hours depending on the model that is running.
• Check the output from the heatmap. Choose the optimizer and learning rate with the best performance.
• Analyse the value and find the values for the learning rate (Eg: 0.01) and Optimizer (Eg: SGM).
• Feed this value back into your model to train the model on these values.

Workflow to train and test the model:

• Train the chosen the chosen model on the given dataset to classify as 'Positive' and 'Negative'. (Note: You need not run grid search again)
• Run the validation and testing cells and analyse your results.
• The validation and training accuracy plot describes how the model performance changes with an increase in epochs.
• The validation and training loss plot describes how the model loss changes with an increase in epochs.
• The confusion matrix and AUC curves help describe the precision and recall on the test dataset.

Testing the model:

• Choose a given image from the dataset and feed into the input layer of the model to find the classification of the particular image.