Fast Sparse 3D Convolution Network with VDB
This addresses the computational bottleneck for 3D object classification tasks, offering a significant speed improvement for applications like robotics or medical imaging.
The paper tackled the problem of inefficient inference for sparse 3D data in convolutional neural networks by proposing a new implementation using NanoVDB, resulting in a 20 times faster performance compared to the state-of-the-art dense CNN model.
We proposed a new Convolution Neural Network implementation optimized for sparse 3D data inference. This implementation uses NanoVDB as the data structure to store the sparse tensor. It leaves a relatively small memory footprint while maintaining high performance. We demonstrate that this architecture is around 20 times faster than the state-of-the-art dense CNN model on a high-resolution 3D object classification network.