Improving The Practicality of Active Learning Pipelines in Real-World Problem Settings: A Case Study in The Classification of Astronomical Data