Advances in Efficient Labeling and Active Learning

The field of machine learning is moving towards more efficient labeling and active learning techniques, with a focus on reducing the need for large amounts of labeled data. Recent developments have shown that selective querying of the most informative samples can significantly improve model performance, while minimizing labeling costs. This is particularly important in applications such as medical imaging, where labeling can be time-consuming and costly. Noteworthy papers in this area include Is Complete Labeling Necessary, which proposes a novel active learning framework for longitudinal medical imaging, and Hierarchical Semi-Supervised Active Learning for Remote Sensing, which integrates semi-supervised learning and active learning for remote sensing applications. Additionally, papers such as StereoDETR and MetaDCSeg have demonstrated the effectiveness of efficient labeling and active learning techniques in 3D object detection and medical image segmentation, respectively. Overall, the field is moving towards more efficient and effective labeling techniques, with a focus on reducing costs and improving model performance.

Sources

Is Complete Labeling Necessary? Understanding Active Learning in Longitudinal Medical Imaging

Hierarchical Semi-Supervised Active Learning for Remote Sensing

StereoDETR: Stereo-based Transformer for 3D Object Detection

MetaDCSeg: Robust Medical Image Segmentation via Meta Dynamic Center Weighting

nnActive: A Framework for Evaluation of Active Learning in 3D Biomedical Segmentation

IDEAL-M3D: Instance Diversity-Enriched Active Learning for Monocular 3D Detection

BackSplit: The Importance of Sub-dividing the Background in Biomedical Lesion Segmentation

How to Purchase Labels? A Cost-Effective Approach Using Active Learning Markets

Self-Paced Learning for Images of Antinuclear Antibodies

Active Learning for GCN-based Action Recognition

Built with on top of