In scientific papers, computer vision models are usually evaluated on well-defined training and test datasets. In practice, however, collecting high-quality data that accurately represents the real world is a challenging problem. Developing models using a non-representative dataset will give high […]