aiqbench analysis
This research addresses a critical bottleneck in 3D perception for autonomous systems by tackling "performance discrepancy" during class-incremental learning across heterogeneous data sources like CAD models and noisy camera scans. The proposed PolyMem method offers an exemplar-free solution that implicitly models high-order feature statistics to improve cross-domain robustness without relying on stored examples. For practitioners building continuous learning pipelines in robotics or AR/VR, this work provides a necessary step toward handling the inherent noise and variability of real-world 3D point clouds.
Mitigating Performance Discrepancy in Cross-Domain 3D Class-Incremental Learning
arXiv:2609.04860v1 Announce Type: cross Abstract: 3D perception plays a crucial role in real-world applications such as autonomous driving, robotics, and AR/VR. In practical scenarios, 3D perception models need to continually adapt to newly emerging 3D object categories, making class-incremental learning (CIL) particularly important. However, unlike 2D images, 3D point clouds are inherently heterogeneous: objects from the same class may not only come from the clean CAD domain, but also from RGB-D camera scans of varying quality, video reconstructions, or even corrupted observations. We discover that such heterogeneity introduces a new challenge beyond catastrophic forgetting: the degree of performance degradation can vary substantially across domains, a phenomenon we term performance discrepancy. To investigate this problem, we establish the Domain3D-CIL training and evaluation protocol, which contains point cloud categories from heterogeneous domains. We further adapt a wide range of mainstream CIL methods to the 3D modality. The results demonstrate that this performance discrepancy consistently appears across these baselines. To mitigate this issue, we introduce PolyMem, an exemplar-free approach that implicitly models rich high-order statistics of the feature distribution to enhance cross-domain robustness. Experiments demonstrate that our method effectively alleviates the performance discrepancy while improving the model's performance across domains. Code will be made publicly available upon acceptance.
Abstract:3D perception plays a crucial role in real-world applications such as autonomous driving, robotics, and AR/VR. In practical scenarios, 3D perception models need to continually adapt to newly emerging 3D object categories, making class-incremental learning (CIL) particularly important. However, unlike 2D images, 3D point clouds are inherently heterogeneous: objects from the same class may not only come from the clean CAD domain, but also from RGB-D camera scans of varying quality, video reconstructions, or even corrupted observations. We discover that such heterogeneity introduces a new challenge beyond catastrophic forgetting: the degree of performance degradation can vary substantially across domains, a phenomenon we term performance discrepancy. To investigate this problem, we establish the Domain3D-CIL training and evaluation protocol, which contains point cloud categories from heterogeneous domains. We further adapt a wide range of mainstream CIL methods to the 3D modality. The results demonstrate that this performance discrepancy consistently appears across these baselines. To mitigate this issue, we introduce PolyMem, an exemplar-free approach that implicitly models rich high-order statistics of the feature distribution to enhance cross-domain robustness. Experiments demonstrate that our method effectively alleviates the performance discrepancy while improving the model's performance across domains. Code will be made publicly available upon acceptance.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2609.04860 [cs.CV]
(or arXiv:2609.04860v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2609.04860 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 300 words
Related coverage
Brain foundation model-guided source-selective domain adaptation for cross-subject EEG decoding
arXiv · Oct 9, 2026
Selective Backpropagation for Efficient Few-Shot Class-Incremental Learning
arXiv · Oct 7, 2026
34 Amazon Research Awards Build on Trainium recipients announced
Amazon Science · Aug 5, 2026
Overcoming Catastrophic Forgetting in Visual Continual Learning with Reinforcement Fine-Tuning
arXiv · Oct 14, 2026