This paper presents a comprehensive survey on Deep Active Learning (DAL), summarizing recent advances and new frontiers in the field. Active learning aims to achieve strong performance with fewer training samples by iteratively selecting the most informative samples for labeling. DAL has gained popularity due to its efficiency in reducing annotation costs, especially in deep learning applications. The paper provides a detailed taxonomy of DAL methods from five perspectives: annotation types, query strategies, deep model architectures, learning paradigms, and training processes. It also summarizes the main applications of DAL in Natural Language Processing (NLP), Computer Vision (CV), and Data Mining (DM), and discusses the challenges and future directions of the field. The paper reviews the most influential DAL baselines and widely used datasets, and discusses the challenges in DAL, including pipeline-related issues, task-related difficulties, and dataset-related problems. It highlights the potential of DAL as a sample selection strategy for large-scale pre-trained models, and the need for a universal framework that is friendly to various downstream tasks. The paper also discusses the integration of DAL with semi-supervised strategies, and the challenges in combining DAL with semi-supervised methods. It emphasizes the importance of scalability and generalizability in DAL, and the need for further research in generative tasks. The paper concludes that DAL has great potential in the field of deep learning, and that further research is needed to address the challenges and opportunities in the field.This paper presents a comprehensive survey on Deep Active Learning (DAL), summarizing recent advances and new frontiers in the field. Active learning aims to achieve strong performance with fewer training samples by iteratively selecting the most informative samples for labeling. DAL has gained popularity due to its efficiency in reducing annotation costs, especially in deep learning applications. The paper provides a detailed taxonomy of DAL methods from five perspectives: annotation types, query strategies, deep model architectures, learning paradigms, and training processes. It also summarizes the main applications of DAL in Natural Language Processing (NLP), Computer Vision (CV), and Data Mining (DM), and discusses the challenges and future directions of the field. The paper reviews the most influential DAL baselines and widely used datasets, and discusses the challenges in DAL, including pipeline-related issues, task-related difficulties, and dataset-related problems. It highlights the potential of DAL as a sample selection strategy for large-scale pre-trained models, and the need for a universal framework that is friendly to various downstream tasks. The paper also discusses the integration of DAL with semi-supervised strategies, and the challenges in combining DAL with semi-supervised methods. It emphasizes the importance of scalability and generalizability in DAL, and the need for further research in generative tasks. The paper concludes that DAL has great potential in the field of deep learning, and that further research is needed to address the challenges and opportunities in the field.