As we did for the Flowers dataset, let’s now conduct the preprocessing steps with the CXR dataset. Due to the nature of the CXR images, some of these steps will be a little different.
For the flowers dataset, we did all the augmentation we possibly could. However, for the CXR data, we had some specific constraints.
- Vertical flip needs to be set as ‘False’. This is because CXR images have a natural orientation – up to down.
- We cannot (i.e. should not) do a centre crop for CXR images, as the anomaly can be in an area outside the cropped portion of the image.
Coming Up
Now that we’re through with augmentation, the next step in data preprocessing is normalisation.
Report an error