A Data-centric Approach to Class-specific Bias in Image Data Augmentation: Appendices A-L | HackerNoon
Briefly

from Hackernoon7 months ago
Data augmentation is a powerful strategy to boost the generalization of models in computer vision. However, it can unintentionally introduce biases, resulting in uneven accuracy across different classes.
Hackernoonhttps://hackernoon.com/a-data-centric-approach-to-class-specific-bias-in-image-data-augmentation-appendices-a-l
While techniques like random cropping and horizontal flipping enhance model robustness, they can affect certain class accuracies detrimentally, showcasing the importance of cautious implementation.
Hackernoonhttps://hackernoon.com/a-data-centric-approach-to-class-specific-bias-in-image-data-augmentation-appendices-a-l
Read at Hackernoon
#data-augmentation #computer-vision #model-generalization #bias-in-ai #class-accuracy
[
|
]