MINER: Multiscale Implicit Neural Representations

Saragadam, Vishwanath; Tan, Jasper; Balakrishnan, Guha; Baraniuk, Richard G.; Veeraraghavan, Ashok

doi:10.48550/arxiv.2202.03532

Cited by 5 publications

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[5,82] utilize meta-learning techniques [20,49] to initialize network parameters, thereby improving the training speed. Some methods [7,10,55,70,80,97] attempt to design scene representations that support efficient training. [21,48,74,83] augments the approximation ability of networks by designing encoding techniques.…”

Section: Related Workmentioning

confidence: 99%

Learning Neural Volumetric Representations of Dynamic Humans in Minutes

Chen¹,

Peng²,

Zhao³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

Section: Related Workmentioning

confidence: 99%

Learning Neural Volumetric Representations of Dynamic Humans in Minutes

Chen¹,

Peng²,

Zhao³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…[Fathony et al 2020] propose a band-limited network to obtain a multi-scale representation by restricting the frequency magnitude of the basis functions. Recently, [Saragadam et al 2022] adopt the Laplacian pyramid to extract multi-scale coefficients for multiple neural networks. Unlike our work, this work overfits each input object with an individual representation for efficient storage and rendering.…”

Section: Related Workmentioning

confidence: 99%

Neural Wavelet-domain Diffusion for 3D Shape Generation

Hui¹,

Li²,

Hu³

et al. 2022

Preprint

View full text Add to dashboard Cite

This paper presents a new approach for 3D shape generation, enabling direct generative modeling on a continuous implicit representation in wavelet domain. Specifically, we propose a compact wavelet representation with a pair of coarse and detail coefficient volumes to implicitly represent 3D shapes via truncated signed distance functions and multi-scale biorthogonal wavelets, and formulate a pair of neural networks: a generator based on the diffusion model to produce diverse shapes in the form of coarse coefficient volumes; and a detail predictor to further produce compatible detail coefficient volumes for enriching the generated shapes with fine structures and details. Both quantitative and qualitative experimental results manifest the superiority of our approach in generating diverse and high-quality shapes with complex topology and structures, clean surfaces, and fine details, exceeding the 3D generation capabilities of the state-of-the-art models. CCS CONCEPTS• Computing methodologies → Shape analysis; Neural networks; Mesh models.

show abstract

“…ACORN [26] and CoordX [20] aim at reducing the number of queries to coordinate-based models with different approaches: ACORN [26] adopts a hierarchical way to decompose the multi-scale coordinates while CoordX [20] designs a split MLP architecture to leverage the locality between input coordinate points. The following MINER [42] improves ACORN via a cross-scale similarity prior.…”

Section: Optimization Of Inrsmentioning

confidence: 99%

E-NeRV: Expedite Neural Video Representation with Disentangled Spatial-Temporal Context

Li¹,

Wang²,

Pi³

et al. 2022

Preprint

View full text Add to dashboard Cite

Recently, the image-wise implicit neural representation of videos, NeRV, has gained popularity for its promising results and swift speed compared to regular pixel-wise implicit representations. However, the redundant parameters within the network structure can cause a large model size when scaling up for desirable performance. The key reason of this phenomenon is the coupled formulation of NeRV, which outputs the spatial and temporal information of video frames directly from the frame index input. In this paper, we propose E-NeRV, which dramatically expedites NeRV by decomposing the image-wise implicit neural representation into separate spatial and temporal context. Under the guidance of this new formulation, our model greatly reduces the redundant model parameters, while retaining the representation ability. We experimentally find that our method can improve the performance to a large extent with fewer parameters, resulting in a more than 8× faster speed on convergence. Code is available at https://github.com/kyleleey/E-NeRV.

show abstract

MINER: Multiscale Implicit Neural Representations

Abstract: Scale 0 (finest) 22 min | 4.4mil. Params. 22 min 17mil params.

Cited by 5 publications

References 12 publications

Learning Neural Volumetric Representations of Dynamic Humans in Minutes

Learning Neural Volumetric Representations of Dynamic Humans in Minutes

Neural Wavelet-domain Diffusion for 3D Shape Generation

E-NeRV: Expedite Neural Video Representation with Disentangled Spatial-Temporal Context

Contact Info

Product

Resources

About