Joint Out-of-Distribution Filtering and Data Discovery Active Learning
🎯 CVPR 2025
Overview
Real-world data is messy — it contains out-of-distribution noise and novel categories waiting to be discovered. Previous active learning methods tackle these challenges separately, but Joda is the first to solve both at the same time.
Joint Out-of-distribution filtering and data Discovery Active learning (Joda) filters out OOD data before selecting the most valuable candidates for labeling, all within a deeply entangled training pipeline that builds a common feature space aligning known and novel categories while separating OOD noise.
Key Contributions
🔗 Joint framework — first method to combine OOD filtering and novel category discovery in active learning
⚡ Highly efficient — no auxiliary models, no training access to the unlabeled pool needed
📊 Extensive evaluation — tested across 18 configurations and 3 metrics, consistently achieving the best accuracy and the strongest class discovery-to-OOD filtering balance
🥇 Outperforms all state-of-the-art competitor approaches
Why It Matters
Labeling data is expensive. Joda makes every label count by intelligently filtering noise and discovering new categories — bringing active learning one step closer to practical, real-world deployment.
As the data demand for deep learning models increases, active learning (AL) becomes essential to strategically select samples for labeling, which maximizes data efficiency and reduces training costs. Real-world scenarios necessitate the consideration of incomplete data knowledge within AL. Prior works address handling out-of-distribution (OOD) data, while another research direction has focused on category discovery. However, a combined analysis of real-world considerations combining AL with out-of-distribution data and category discovery remains unexplored. To address this gap, we propose Joint Out-of-distribution filtering and data Discovery Active learning (Joda) , to uniquely address both challenges simultaneously by filtering out OOD data before selecting candidates for labeling. In contrast to previous methods, we deeply entangle the training procedure with filter and selection to construct a common feature space that aligns known and novel categories while separating OOD samples. Unlike previous works, Joda is highly efficient and completely omits auxiliary models and training access to the unlabeled pool for filtering or selection. In extensive experiments on 18 configurations and 3 metrics, \ours{} consistently achieves the highest accuracy with the best class discovery to OOD filtering balance compared to state-of-the-art competitor approaches.
@inproceedings{Schmidt2025a,author={Schmidt, Sebastian and Schenk, Leonard and Schwinn, Leo and Günnemann, Stephan},booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},title={Joint Out-of-Distribution Filtering and Data Discovery Active Learning},url={http://arxiv.org/abs/2503.02491},year={2025},}