Most 3D instance segmentation methods exploit a bottom-up strategy, typically including resource-exhaustive post-processing. For point grouping, bottom-up methods rely on prior assumptions about the objects in the form of hyperparameters, which are domain-specific and need to be carefully tuned. On the contrary, we address 3D instance segmentation with a TD3D: top-down, fully data-driven, simple approach trained in an end-to-end manner. With its straightforward fully-convolutional pipeline, it performs surprisingly well on the standard benchmarks: ScanNet v2, its extension ScanNet200, and S3DIS. Besides, our method is much faster on inference than the current state-of-the-art grouping-based approaches. Code is available at https://github.com/SamsungLabs/td3d .

翻译：大多数3D实例分割方法采用自下而上的策略，通常包含计算资源密集的后处理步骤。在点云分组中，自下而上方法依赖于以超参数形式呈现的关于物体的先验假设，这些超参数具有领域特异性且需要精细调优。相反，我们提出采用TD3D（自上而下、完全数据驱动、简单易行的方法）来处理3D实例分割，该方法以端到端方式训练。凭借其简洁的全卷积流水线，该方法在标准基准数据集（ScanNet v2、其扩展集ScanNet200及S3DIS）上取得了出人意料的优异表现。此外，在推理速度上，我们的方法远快于当前最先进的基于分组的方法。代码已开源：https://github.com/SamsungLabs/td3d。