HollowNeRF: Pruning Hashgrid-Based NeRFs with Trainable Collision Mitigation
This work addresses the problem of inefficient parameter usage in NeRF for 3D scene rendering, offering a compression solution that improves cost-accuracy trade-offs, though it is incremental as it builds on existing hashgrid-based methods.
The paper tackles the challenge of leveraging spatial sparsity in 3D scenes for hashgrid-based Neural Radiance Fields (NeRF) by proposing HollowNeRF, which automatically sparsifies the feature grid during training using a coarse 3D saliency mask and ADMM pruner, resulting in comparable rendering quality to Instant-NGP with only 31% of the parameters and up to 1dB PSNR gain with 56% of parameters.
Neural radiance fields (NeRF) have garnered significant attention, with recent works such as Instant-NGP accelerating NeRF training and evaluation through a combination of hashgrid-based positional encoding and neural networks. However, effectively leveraging the spatial sparsity of 3D scenes remains a challenge. To cull away unnecessary regions of the feature grid, existing solutions rely on prior knowledge of object shape or periodically estimate object shape during training by repeated model evaluations, which are costly and wasteful. To address this issue, we propose HollowNeRF, a novel compression solution for hashgrid-based NeRF which automatically sparsifies the feature grid during the training phase. Instead of directly compressing dense features, HollowNeRF trains a coarse 3D saliency mask that guides efficient feature pruning, and employs an alternating direction method of multipliers (ADMM) pruner to sparsify the 3D saliency mask during training. By exploiting the sparsity in the 3D scene to redistribute hash collisions, HollowNeRF improves rendering quality while using a fraction of the parameters of comparable state-of-the-art solutions, leading to a better cost-accuracy trade-off. Our method delivers comparable rendering quality to Instant-NGP, while utilizing just 31% of the parameters. In addition, our solution can achieve a PSNR accuracy gain of up to 1dB using only 56% of the parameters.