The scenario
A fall detection dataset from wearable accelerometers has too few fall samples; the pipeline assesses imbalance and applies data augmentation before training.
What is in this drawing
Read the decisions behind it.
01
Apply augmentation only when class distribution is imbalanced
02
Validate synthetic sample quality and refine parameters if poor
03
Merge augmented data with real data and reassess balance
Reusable for time-series classification tasks with imbalanced classes, especially wearable sensor data.
More like this