Interpolation-Aware Padding for 3D Sparse Convolutional Neural Networks
This addresses a specific bottleneck in fine-grained 3D vision tasks, offering an incremental improvement over current padding schemes.
The paper tackles the problem of inaccurate point-wise feature computation in 3D sparse CNNs for tasks like semantic segmentation and 3D detection by proposing interpolation-aware padding, which pads empty voxels near non-empty ones to enable trilinear interpolation, resulting in higher prediction accuracy compared to existing methods.
Sparse voxel-based 3D convolutional neural networks (CNNs) are widely used for various 3D vision tasks. Sparse voxel-based 3D CNNs create sparse non-empty voxels from the 3D input and perform 3D convolution operations on them only. We propose a simple yet effective padding scheme --- interpolation-aware padding to pad a few empty voxels adjacent to the non-empty voxels and involve them in the 3D CNN computation so that all neighboring voxels exist when computing point-wise features via the trilinear interpolation. For fine-grained 3D vision tasks where point-wise features are essential, like semantic segmentation and 3D detection, our network achieves higher prediction accuracy than the existing networks using the nearest neighbor interpolation or the normalized trilinear interpolation with the zero-padding or the octree-padding scheme. Through extensive comparisons on various 3D segmentation and detection tasks, we demonstrate the superiority of 3D sparse CNNs with our padding scheme in conjunction with feature interpolation.