Introduction

In the previous session, you learnt how to set up a pipeline for building and training CNNs. In this session, you will apply your skills to detect anomalies in chest X-Ray scans.

Recap of Techniques Learnt

You performed the following steps in the previous session:

1. Data Preparation:
   1. Made sure all our images were of the same resolution.
   2. Placed the images in two different folders – ‘rose’ and ‘daisy’. This method will work
   for any application where you ‘re trying to train using images.
2. Data Pre-processing: Morphological Operations
   1. Did thresholding on the image – converted it from a grey image to a binary image.
   2. Looked at Erosion, Dilation, Opening, Closing.
3. Data Pre-Processing: Augmentation.
   1. Understood the need for data augmentation.
   2. Saw some commonly used methods of normalisation.
4. Data Pre-Processing: Augmentation
   1. Understood the need for data augmentation.
   2. Learnt about two types of transformations for augmentation – linear and affine.
   3. Saw different ways to augment – translation, rotation, scaling, etc.
5. Model Building
   1. Running ablation experiments
   2. Overfitting on a smaller version of the training set
   3. Hyperparameter tuning
   4. Mode training and evaluation

Application to Chest X- rays

In this session, you will use these above methodologies from the previous session onto Chest X-ray data images.

People You Will Hear From In This Session

Rohit Ghosh

AI Researcher, Qure.ai

Report an error