S3IM: Stochastic Structural SIMilarity and Its Unreasonable Effectiveness for Neural Fields

• URL: http://arxiv.org/abs/2308.07032v1
• Date: Mon, 14 Aug 2023 09:45:28 GMT
• Title: S3IM: Stochastic Structural SIMilarity and Its Unreasonable Effectiveness for Neural Fields
• Authors: Zeke Xie, Xindi Yang, Yujie Yang, Qi Sun, Yixiang Jiang, Haoran Wang, Yunfeng Cai, Mingming Sun
• Abstract summary: We show that Structural SIMilarity (S3IM) loss processes multiple data points as a whole set instead of multiplexing multiple inputs independently. Our experiments demonstrate the unreasonable effectiveness of S3IM in improving NeRF and neural surface representation for nearly free.
• Score: 46.9880016170926
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recently, Neural Radiance Field (NeRF) has shown great success in rendering novel-view images of a given scene by learning an implicit representation with only posed RGB images. NeRF and relevant neural field methods (e.g., neural surface representation) typically optimize a point-wise loss and make point-wise predictions, where one data point corresponds to one pixel. Unfortunately, this line of research failed to use the collective supervision of distant pixels, although it is known that pixels in an image or scene can provide rich structural information. To the best of our knowledge, we are the first to design a nonlocal multiplex training paradigm for NeRF and relevant neural field methods via a novel Stochastic Structural SIMilarity (S3IM) loss that processes multiple data points as a whole set instead of process multiple inputs independently. Our extensive experiments demonstrate the unreasonable effectiveness of S3IM in improving NeRF and neural surface representation for nearly free. The improvements of quality metrics can be particularly significant for those relatively difficult tasks: e.g., the test MSE loss unexpectedly drops by more than 90% for TensoRF and DVGO over eight novel view synthesis tasks; a 198% F-score gain and a 64% Chamfer $L_{1}$ distance reduction for NeuS over eight surface reconstruction tasks. Moreover, S3IM is consistently robust even with sparse inputs, corrupted images, and dynamic scenes.

Related papers

• MomentsNeRF: Leveraging Orthogonal Moments for Few-Shot Neural Rendering [4.6786468967610055]
  We propose MomentsNeRF, a novel framework for one- and few-shot neural rendering. Our architecture offers a new transfer learning method to train on multi-scenes. Our approach is the first to successfully harness features extracted from Gabor and Zernike moments.
  arXiv  Detail & Related papers  (2024-07-02T21:02:48Z)
• NeRF-DetS: Enhancing Multi-View 3D Object Detection with Sampling-adaptive Network of Continuous NeRF-based Representation [60.47114985993196]
  NeRF-Det unifies the tasks of novel view arithmetic and 3D perception. We introduce a novel 3D perception network structure, NeRF-DetS. NeRF-DetS outperforms competitive NeRF-Det on the ScanNetV2 dataset.
  arXiv  Detail & Related papers  (2024-04-22T06:59:03Z)
• Leveraging Neural Radiance Fields for Uncertainty-Aware Visual Localization [56.95046107046027]
  We propose to leverage Neural Radiance Fields (NeRF) to generate training samples for scene coordinate regression. Despite NeRF's efficiency in rendering, many of the rendered data are polluted by artifacts or only contain minimal information gain.
  arXiv  Detail & Related papers  (2023-10-10T20:11:13Z)
• Learning Neural Duplex Radiance Fields for Real-Time View Synthesis [33.54507228895688]
  We propose a novel approach to distill and bake NeRFs into highly efficient mesh-based neural representations. We demonstrate the effectiveness and superiority of our approach via extensive experiments on a range of standard datasets.
  arXiv  Detail & Related papers  (2023-04-20T17:59:52Z)
• AligNeRF: High-Fidelity Neural Radiance Fields via Alignment-Aware Training [100.33713282611448]
  We conduct the first pilot study on training NeRF with high-resolution data. We propose the corresponding solutions, including marrying the multilayer perceptron with convolutional layers. Our approach is nearly free without introducing obvious training/testing costs.
  arXiv  Detail & Related papers  (2022-11-17T17:22:28Z)
• Neural Knitworks: Patched Neural Implicit Representation Networks [1.0470286407954037]
  We propose Knitwork, an architecture for neural implicit representation learning of natural images that achieves image synthesis. To the best of our knowledge, this is the first implementation of a coordinate-based patch tailored for synthesis tasks such as image inpainting, super-resolution, and denoising. The results show that modeling natural images using patches, rather than pixels, produces results of higher fidelity.
  arXiv  Detail & Related papers  (2021-09-29T13:10:46Z)
• UNISURF: Unifying Neural Implicit Surfaces and Radiance Fields for Multi-View Reconstruction [61.17219252031391]
  We present a novel method for reconstructing surfaces from multi-view images using Neural implicit 3D representations. Our key insight is that implicit surface models and radiance fields can be formulated in a unified way, enabling both surface and volume rendering. Our experiments demonstrate that we outperform NeRF in terms of reconstruction quality while performing on par with IDR without requiring masks.
  arXiv  Detail & Related papers  (2021-04-20T15:59:38Z)
• MVSNeRF: Fast Generalizable Radiance Field Reconstruction from Multi-View Stereo [52.329580781898116]
  We present MVSNeRF, a novel neural rendering approach that can efficiently reconstruct neural radiance fields for view synthesis. Unlike prior works on neural radiance fields that consider per-scene optimization on densely captured images, we propose a generic deep neural network that can reconstruct radiance fields from only three nearby input views via fast network inference.
  arXiv  Detail & Related papers  (2021-03-29T13:15:23Z)
• pixelNeRF: Neural Radiance Fields from One or Few Images [20.607712035278315]
  pixelNeRF is a learning framework that predicts a continuous neural scene representation conditioned on one or few input images. We conduct experiments on ShapeNet benchmarks for single image novel view synthesis tasks with held-out objects. In all cases, pixelNeRF outperforms current state-of-the-art baselines for novel view synthesis and single image 3D reconstruction.
  arXiv  Detail & Related papers  (2020-12-03T18:59:54Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.