解决3D感知中新增类别时数据污染导致遗忘的问题
Robust3D-CIL: Robust Class-Incremental Learning for 3D Perception
- 用最远点采样思想选样本，保持类别内多样性
- 通过点云下采样优化存储，提升记忆效率
- 在真实污染数据上性能提升2%-11%，适合自动驾驶等场景
3D感知在自动驾驶、机器人和AR/VR等实际应用中至关重要。现实中模型需持续适应新数据与新类别，但重新训练成本过高，因此类增量学习（CIL）尤为关键。然而，真实3D点云常含噪声数据，现有CIL方法在污染数据上易产生严重遗忘。为此，本文研究在未知污染条件下更新模型的场景。受最远点采样启发，提出一种新原型选择策略，有效保留类别内多样性，缓解数据污染引发的遗忘。同时引入基于点云下采样的重放方法，更高效利用有限的重放缓冲区内存，增强模型持续学习能力。大量实验表明，该方法使基于重放的CIL基线性能提升2%至11%，验证了其有效性及在真实3D应用中的潜力。
原文摘要 · Abstract (English)
3D perception plays a crucial role in real-world applications such as autonomous driving, robotics, and AR/VR. In practical scenarios, 3D perception models must continuously adapt to new data and emerging object categories, but retraining from scratch incurs prohibitive costs. Therefore, adopting class-incremental learning (CIL) becomes particularly essential. However, real-world 3D point cloud data often include corrupted samples, which poses significant challenges for existing CIL methods and leads to more severe forgetting on corrupted data. To address these challenges, we consider the scenario in which a CIL model can be updated using point clouds with unknown corruption to better simulate real-world conditions. Inspired by Farthest Point Sampling, we propose a novel exemplar selection strategy that effectively preserves intra-class diversity when selecting replay exemplars, mitigating forgetting induced by data corruption. Furthermore, we introduce a point cloud downsampling-based replay method to utilize the limited replay buffer memory more efficiently, thereby further enhancing the model's continual learning ability. Extensive experiments demonstrate that our method improves the performance of replay-based CIL baselines by 2% to 11%, proving its effectiveness and promising potential for real-world 3D applications.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。