Backpropagation-free method improves 3D point cloud recognition under domain shifts
3D-BBDA: Backpropagation-Free Black-Box Domain Adaptation for Robust 3D Point Cloud Recognition
Computer Vision and Pattern RecognitionComputational Engineering, Finance, and ScienceRobotics
Summary
3D point cloud recognition systems often perform worse when the data they see changes, like from different sensors or environments. The authors developed a new method that improves these systems without needing to change the original model or access its internal details, which is helpful when data or models are private. Their approach uses clever ways to guess labels and align features using only the model's output, making it faster and easier to use in real-time. Tests show their method works better than strong previous techniques on several challenging datasets.
What this means in practice
- •For robotics developers: Enhance robot perception systems to handle variations in sensor data without retraining or accessing original training data.
- •For mapping engineers: Improve 3D urban mapping accuracy when environmental and sensor conditions differ from original model training setups.
Authors
Shivangi Rai, Kunal Jangid, Vaibhav Kumar, Vinod K Kurmi
Abstract
Robust 3D point-cloud recognition is critical for robotics, mapping, and urban sensing, yet point-cloud classifiers often suffer performance drops under domain shifts caused by sensor changes, environmental variation, and acquisition artifacts. Many existing adaptation methods are difficult to deploy in practice because they require access to source data or the internal parameters of the source model, which may be restricted by privacy, licensing, or security constraints. We present 3D-BBDA, a backpropagation-free black-box domain adaptation framework that adapts a source-trained predictor to unlabeled target data using only black-box outputs. Our method performs pseudo-label guided prototype-based alignment and employs a weighted prediction module to fuse source logits with alignment logits, reducing sensitivity to noisy pseudo-labels. To stabilize prototype memory when pseudo-label coverage is incomplete, we construct a Gaussian-perturbed prototype prior from unlabeled target predictions and use it as a fallback for missing or unreliable class prototypes during alignment. Experiments on four corrupted 3D benchmarks demonstrate consistent improvements over strong baselines. On ModelNet-40C, 3D-BBDA achieves mean errors of 32.08% (PointNet) and 29.62% (GDANet), outperforming Source-only (35.88% / 34.03%) and BFTT3D (33.78% / 32.36%). 3D-BBDA requires only a single forward pass with no gradient updates making it suitable for real-time, privacy-preserving deployment in transferable 3D recognition pipelines. Project page is available at https://visdomlab.github.io/3D_BBDA