Augmentation on the How to Train Models track. Augmentation invents extra training inputs that should not change the label. A flip is fine for a cat. It is wrong for a digit 6 if it becomes a 9. Never augment the test set.
This lesson assumes you already worked through Imbalanced classes.
The idea in practice
Apply augmentation on train only, in the loader. Document each transform.
A concrete check
goal = {
'track': 'How to Train Models',
'lesson': 'Augmentation',
}
checks = [
'input available at decision time',
'score matches the real decision',
'failure case written down',
]
print(goal['lesson'])
for item in checks:
print('-', item)
Run the sketch locally if you have Python. The printout is a reminder of the checks, not a trained model. Replace the strings with the real inputs from your own example before you treat it as a design.
What usually goes wrong
Augmenting validation so the score measures your transforms, not the real data. When this happens, stop adding parameters or tools. Fix the check, the data, or the permission, then run the same example again.
What to write down
- The input you are allowed to use at decision time.
- The output and the score or pass rule.
- One failure you will test on purpose.
- What you will not claim the system can do.
Practice
Name one augmentation that would change a label in your task.
Self-check
- Say Augmentation in one sentence that mentions an input and an output.
- Name the failure mode in this lesson and the check that would catch it.
Done when: you can explain this lesson without the page open, and you have a written failure case.