Putting 3D Spatially Sparse Networks on a Diet
This work addresses efficiency issues in 3D vision for applications like indoor and outdoor segmentation, though it is incremental as it builds on existing pruning methods.
The paper tackles the problem of inefficient parameter usage in 3D neural networks for vision tasks by proposing a weight-sparse and spatially sparse 3D convolutional network (WS^3-Convnet), achieving minimal performance loss (2.15% drop) with significant reductions in parameters (99% compression) and computational cost (95% reduction).
3D neural networks have become prevalent for many 3D vision tasks including object detection, segmentation, registration, and various perception tasks for 3D inputs. However, due to the sparsity and irregularity of 3D data, custom 3D operators or network designs have been the primary focus of research, while the size of networks or efficacy of parameters has been overlooked. In this work, we perform the first comprehensive study on the weight sparsity of spatially sparse 3D convolutional networks and propose a compact weight-sparse and spatially sparse 3D convnet (WS^3-Convnet) for semantic and instance segmentation on the real-world indoor and outdoor datasets. We employ various network pruning strategies to find compact networks and show our WS^3-Convnet achieves minimal loss in performance (2.15\% drop) with orders-of-magnitude smaller number of parameters (99\% compression rate) and computational cost (95\% reduction). Finally, we systematically analyze the compression patterns of WS^3-Convnet and show interesting emerging sparsity patterns common in our compressed networks to further speed up inference (45\% faster). \keywords{Efficient network architecture, Network pruning, 3D scene segmentation, Spatially sparse convolution}