Volume upscaling with convolutional neural networks

Zhou, Zhenglei; Hou, Yule; Wang, Qirui; Chen, Guangxiang; Lu, Jiawei; Tao, Yubo; Lin, Hai

doi:10.1145/3095140.3095178

Cited by 34 publications

(22 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Spatial super resolution aims to increase the spatial resolution of input data. Zhou et al [73] use a 3D convolutional neural network (CNN) to perform SSR on volumes with better feature reconstruction than trilinear interpolation or cubic-spline interpolation. Guo et al [18] use three neural networks in parallel to perform SSR on 3D vector fields.…”

Section: Super Resolution For Scientific Datamentioning

confidence: 99%

“…Interpolation techniques like linear or bicubic interpolation are used to fill in the discarded data post-hoc, but can lead to overly smooth results and missing high-frequency features important for analysis. Due to the recent advancements of machine learning (ML), super resolution (SR) methods applied to scientific data have seen better feature re-construction than interpolation methods [18][19][20]73]. Using neural networks (NNs) for non-error-bounded lossy compression has been studied extensively for image compression with results exceeding JPEG and JPEG2000 [1,5,55,[57][58][59].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Hierarchical Super Resolution for Scientific Data

Wurster¹,

Guo²,

Shen³

et al. 2021

Preprint

View full text Add to dashboard Cite

We present an approach for hierarchical super resolution (SR) using neural networks on an octree data representation. We train a hierarchy of neural networks, each capable of 2× upscaling in each spatial dimension between two levels of detail, and use these networks in tandem to facilitate large scale factor super resolution, scaling with the number of trained networks. We utilize these networks in a hierarchical super resolution algorithm that upscales multiresolution data to a uniform high resolution without introducing seam artifacts on octree node boundaries. We evaluate application of this algorithm in a data reduction framework by dynamically downscaling input data to an octree-based data structure to represent the multiresolution data before compressing for additional storage reduction. We demonstrate that our approach avoids seam artifacts common to multiresolution data formats, and show how neural network super resolution assisted data reduction can preserve global features better than compressors alone at the same compression ratios.

show abstract

Section: Super Resolution For Scientific Datamentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Deep Hierarchical Super Resolution for Scientific Data

Wurster¹,

Guo²,

Shen³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Also in scientific data visualization researchers have begun to explore the capabilities of CNNs for upscaling and reconstruction of 2D/3D steady and time-varying scientific data, including both scalar and vector fields. Zhou et al (2017) presented a CNN-based solution that downscales a volumetric dataset using three hidden layers designed for feature extraction, nonlinear mapping and reconstruction, respectively. Han et al (2019) took a twostage approach for vector field reconstruction via deep learning.…”

Section: Upscaling Of Images and Physical Fieldsmentioning

confidence: 99%

A comparative study of convolutional neural network models for wind field downscaling

Höhlein

Kern

Hewson

et al. 2020

Meteorological Applications

View full text Add to dashboard Cite

We analyze the applicability of convolutional neural network (CNN) architectures for downscaling of short‐range forecasts of near‐surface winds on extended spatial domains. Short‐range wind forecasts (at the 100 m level) from European Centre for Medium Range Weather Forecasts ERA5 reanalysis initial conditions at 31 km horizontal resolution are downscaled to mimic high resolution (HRES) (deterministic) short‐range forecasts at 9 km resolution. We evaluate the downscaling quality of four exemplary CNN architectures and compare these against a multilinear regression model. We conduct a qualitative and quantitative comparison of model predictions and examine whether the predictive skill of CNNs can be enhanced by incorporating additional atmospheric variables, such as geopotential height and forecast surface roughness, or static high‐resolution fields, like land–sea mask and topography. We further propose DeepRU, a novel U‐Net‐based CNN architecture, which is able to infer situation‐dependent wind structures that cannot be reconstructed by other models. Inferring a target 9 km resolution wind field from the low‐resolution input fields over the Alpine area takes less than 10 ms on our graphics processing unit target architecture, which compares favorably to an overhead in simulation time of minutes or hours between low‐ and high‐resolution forecast simulations.

show abstract

“…In visualization, Zhou et al [64] presented a CNN-based solution that upscales a volumetric data set using three hidden layers designed for feature extraction, non-linear mapping, and reconstruction, respectively. Han et al [20] introduced a two-stage approach for vector field reconstruction via deep learning, by refining a low-resolution vector field from a set of streamlines.…”

Section: Related Workmentioning

confidence: 99%

“…to infer missing data samples. This type of reconstruction has been performed in the visualization image domain to infer highresolution images from given low-resolution images of isosurfaces [60], in the spatial domain to infer a higher resolution of a 3D data set from a low-resolution version [64], and in the temporal domain to infer a temporally dense volume sequence from a sparse temporal sequence [21].…”

Section: Introductionmentioning

confidence: 99%

Learning Adaptive Sampling and Reconstruction for Volume Visualization

Weiss,

Işık,

Thies

et al. 2020

Preprint

View full text Add to dashboard Cite

Volume upscaling with convolutional neural networks

Cited by 34 publications

References 10 publications

Deep Hierarchical Super Resolution for Scientific Data

Deep Hierarchical Super Resolution for Scientific Data

A comparative study of convolutional neural network models for wind field downscaling

Learning Adaptive Sampling and Reconstruction for Volume Visualization

Contact Info

Product

Resources

About