Report

• https://github.com/arrouasserudy/chest-anomalies-detection/blob/main/Report%20Chest%20X-ray%20Abnormalities%20Detection.pdf

How to use this repo

• Install the dependencies in requirement.txt

• Download the data at https://www.kaggle.com/xhlulu/vinbigdata-chest-xray-resized-png-256x256 and unzip into the data/ folder. The original data not resized can be dowloaded here : https://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/data

• Run the main.py file with the action parameter in this way: python main.py <action>

• The action parameter can be one of the following:

  • init - Create the full metadata file needed to train the model
  • train - Train the model with the parameters setted in 'params' in the main.py file
  • eval - Compute metrics (APs and mAP) for each model saved in checkpoints/
  • loss - Show the graph loss/epochs for each model saved in checkpoints/
  • predict - Show examples of predictions using the model given as a param in the main_predict function

Result examples: