🤖 AI Summary
This work addresses the problem that boundary-region misclassifications in 3D semantic segmentation are obscured by mainstream evaluation metrics. Methodologically, we propose a fine-grained error taxonomy and four boundary-sensitive evaluation metrics; introduce the first boundary–semantic feature disentanglement module; design an attention-guided fusion mechanism; and develop a lightweight boundary pseudo-labeling algorithm—enabling 3.9× inference acceleration and compatibility with standard data augmentation. Our contributions are threefold: (1) We systematically define and optimize boundary-related error metrics for the first time, advancing segmentation quality toward structural plausibility and geometric robustness; (2) We achieve significant improvements in boundary-specific metrics across major benchmarks while maintaining state-of-the-art performance on general metrics such as mIoU; and (3) The method is computationally efficient and real-time capable, with open-sourced implementation.
📝 Abstract
3D semantic segmentation plays a fundamental and crucial role to understand 3D scenes. While contemporary state-of-the-art techniques predominantly concentrate on elevating the overall performance of 3D semantic segmentation based on general metrics (e.g. mIoU, mAcc, and oAcc), they unfortunately leave the exploration of challenging regions for segmentation mostly neglected. In this paper, we revisit 3D semantic segmentation through a more granular lens, shedding light on subtle complexities that are typically overshadowed by broader performance metrics. Concretely, we have delineated 3D semantic segmentation errors into four comprehensive categories as well as corresponding evaluation metrics tailored to each. Building upon this categorical framework, we introduce an innovative 3D semantic segmentation network called BFANet that incorporates detailed analysis of semantic boundary features. First, we design the boundary-semantic module to decouple point cloud features into semantic and boundary features, and fuse their query queue to enhance semantic features with attention. Second, we introduce a more concise and accelerated boundary pseudo-label calculation algorithm, which is 3.9 times faster than the state-of-the-art, offering compatibility with data augmentation and enabling efficient computation in training. Extensive experiments on benchmark data indicate the superiority of our BFANet model, confirming the significance of emphasizing the four uniquely designed metrics. Code is available at https://github.com/weiguangzhao/BFANet.