Locating and Mitigating Gradient Conflicts in Point Cloud Domain Adaptation via Saliency Map Skewness

2025-04-22

AI Summary

In unsupervised domain adaptation (UDA) for point clouds, multi-task learning suffers from gradient conflict—where gradients from auxiliary self-supervised tasks degrade the performance of the primary classification task. Method: To address this, we propose a sample selection mechanism based on the skewness of 3D saliency maps—the first to quantify gradient conflict via saliency map skewness. We design SM-DSB, an unsupervised, plug-and-play module that dynamically identifies and filters out gradient-harmful samples during backpropagation. Contribution/Results: SM-DSB requires no target-domain labels and is compatible with mainstream multi-task UDA frameworks. On benchmarks such as ScanObjectNN → ModelNet, it surpasses state-of-the-art methods, improving average classification accuracy by 2.1–3.8% while incurring less than 5% additional computational overhead. This work establishes a novel, interpretable, and low-cost optimization paradigm for point cloud UDA.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Search and Optimization: Sampling/Simulation-based Search
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Object classification models utilizing point cloud data are fundamental for 3D media understanding, yet they often struggle with unseen or out-of-distribution (OOD) scenarios. Existing point cloud unsupervised domain adaptation (UDA) methods typically employ a multi-task learning (MTL) framework that combines primary classification tasks with auxiliary self-supervision tasks to bridge the gap between cross-domain feature distributions. However, our further experiments demonstrate that not all gradients from self-supervision tasks are beneficial and some may negatively impact the classification performance. In this paper, we propose a novel solution, termed Saliency Map-based Data Sampling Block (SM-DSB), to mitigate these gradient conflicts. Specifically, our method designs a new scoring mechanism based on the skewness of 3D saliency maps to estimate gradient conflicts without requiring target labels. Leveraging this, we develop a sample selection strategy that dynamically filters out samples whose self-supervision gradients are not beneficial for the classification. Our approach is scalable, introducing modest computational overhead, and can be integrated into all the point cloud UDA MTL frameworks. Extensive evaluations demonstrate that our method outperforms state-of-the-art approaches. In addition, we provide a new perspective on understanding the UDA problem through back-propagation analysis.

Problem

• Identifying harmful gradients in point cloud domain adaptation
• Mitigating gradient conflicts via saliency map skewness analysis
• Improving classification in unseen 3D data scenarios

Innovation

• Saliency Map-based Data Sampling Block (SM-DSB)
• Skewness of 3D saliency maps for gradient conflicts
• Dynamic sample selection to filter harmful gradients