Data augmentation is a process that enables you to increase the amount of training data by making reasonable modifications in your existing data. It helps you to increase the diversity of your training data which is essential for developing a robust model. This then, generally speaking, improves the performance of deep learning models.
Although data augmentation can be applied in various domains, it’s commonly used in computer vision. Some of the most common data augmentation techniques used for images are:
- Scaling
- Cropping
- Flipping
- Padding
- Rotation
- Brightness
- Contrast
Why do we need data augmentation?
Development in the field of deep learning has been largely driven by the quality and quantity of data collected and analysed in recent years. Collection of data and then labelling it is a time-consuming activity that can take a considerable amount of time to complete. Sometimes it may become expensive due to the fact that experts in the fields are often involved with the work. Every process associated with data involves a cost in terms of time, money, human effort and computational resources associated with it. Hence we may need to use data augmentation to increase the size of our dataset by applying some transformations to the original data. This helps us to easily gather more data and reduce the cost involved in processing the data.
Advantages of data augmentation
Data augmentation has the following advantages :
- It helps you to generate more data from a limited amount of data. This data can be in a different orientation, location, scale, brightness, etc, which helps the model to become more efficient.
- It helps you to avoid the problem of overfitting. Overfitting occurs as the models learn the whole dataset instead of generalizing the main concepts from it. This means that the model will not be efficient and robust in while making predictions.