|
Dataset and the need for augmentation |
Data! Data! Data! I cannot make bricks without clay - Sherlock Holmes
A model can only be as good as the data we give it. Following the Garbage in, Garbage out policy poor data will severely affect the performance of your model. This effect is more pronounced if you deploy your model in the real world where partial occlusion and illumination variance are severe. While TensorFlow provides certain pre-processing steps that you can use to augment your data we use Roboflow as it made the process for annotation, pre-processing and augmentation extremely streamlined along with providing a detailed analysis of the data. The following are screenshots from the site demonstrating few functionalities of Roboflow.
Back to Main