论文标题
MVTN：3D形状识别的多视图转换网络
MVTN: Multi-View Transformation Network for 3D Shape Recognition
论文作者
论文摘要
多视图投影方法证明了他们在3D形状识别上达到最先进的性能的能力。这些方法学习不同的方法来汇总多个视图的信息。但是，这些视图的摄像机视图往往是针对所有形状的启发式设置和固定的。为了避免缺乏当前多视图方法的活力，我们建议学习这些观点。特别是，我们介绍了多视图转换网络（MVTN），该网络为3D形状识别回归最佳视图，并基于可区分渲染的进步。结果，可以将MVTN与任何用于3D形状分类的多视图网络一起训练。我们将MVTN集成在新型的自适应多视图管道中，该管道可以呈现3D网格或点云。 MVTN在3D形状分类和3D形状检索任务中表现出明显的性能增长，而无需额外的训练监督。在这些任务中，MVTN在ModelNet40，Shapenet Core55以及最新且现实的Scanobjectnn数据集（最高6％）上实现了最先进的性能。有趣的是，我们还表明MVTN可以在3D域中提供针对旋转和遮挡的网络鲁棒性。该代码可在https://github.com/ajhamdi/mvtn上找到。
Multi-view projection methods have demonstrated their ability to reach state-of-the-art performance on 3D shape recognition. Those methods learn different ways to aggregate information from multiple views. However, the camera view-points for those views tend to be heuristically set and fixed for all shapes. To circumvent the lack of dynamism of current multi-view methods, we propose to learn those view-points. In particular, we introduce the Multi-View Transformation Network (MVTN) that regresses optimal view-points for 3D shape recognition, building upon advances in differentiable rendering. As a result, MVTN can be trained end-to-end along with any multi-view network for 3D shape classification. We integrate MVTN in a novel adaptive multi-view pipeline that can render either 3D meshes or point clouds. MVTN exhibits clear performance gains in the tasks of 3D shape classification and 3D shape retrieval without the need for extra training supervision. In these tasks, MVTN achieves state-of-the-art performance on ModelNet40, ShapeNet Core55, and the most recent and realistic ScanObjectNN dataset (up to 6% improvement). Interestingly, we also show that MVTN can provide network robustness against rotation and occlusion in the 3D domain. The code is available at https://github.com/ajhamdi/MVTN .
