pith. sign in

arxiv: 2606.30647 · v1 · pith:2TEIAA4Xnew · submitted 2026-05-31 · 💻 cs.CV · cs.AI

Cross-Modal Hierarchical Fusion for from Multi-Sensor Ground Observation

Xinze Zhang This is my paper

Pith reviewed 2026-07-01 07:17 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords cloud microphysicsmulti-modal fusion4D reconstructionsky cameracloud radarwind estimationvariational refinementground-based observation
0
0 comments X

The pith

AtmoFuseNet fuses sky camera imagery with radar and ceilometer data to produce 4D cloud microphysical fields and wind vectors.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents a three-stage method to reconstruct detailed cloud volumes and motion from heterogeneous ground instruments that each capture only partial information. A cross-modal aggregation step merges image pyramids with vertical profiles using layer-wise attention, a variational module refines the result to satisfy differentiable forward models of radar and imaging, and a correlation step extracts 3D wind from successive volumes. On observations from one semi-arid site the resulting fields reach 0.026 g m^{-3} liquid water content error and 1.18 m s^{-1} wind speed error, beating prior retrieval approaches. A reader would care because the work shows how sparse surface sensors can be combined into time-resolved three-dimensional cloud maps without requiring dense instrument arrays.

Core claim

AtmoFuseNet produces 4D estimates of cloud state and wind by first applying cross-modal hierarchical aggregation to combine image feature pyramids with instrument-derived vertical profiles through layer-wise cross-attention, then using a conditional variational refinement module to map the volume to physically consistent microphysical fields under differentiable radar and image forward models, and finally applying a correlation-based motion estimator to recover per-voxel 3D wind vectors; on collocated observations from a semi-arid site this yields 0.026 g m^{-3} liquid water content MAE and 1.18 m s^{-1} wind speed MAE while outperforming existing baselines, with ablations confirming the con

What carries the argument

The cross-modal hierarchical aggregation module that merges image feature pyramids with vertical profiles via layer-wise cross-attention, followed by conditional variational refinement under differentiable forward models.

If this is right

  • Sparse ground instruments can support dense volumetric cloud reconstruction in both space and time.
  • Per-voxel three-dimensional wind vectors become recoverable from consecutive reconstructed volumes.
  • Each of the three stages contributes measurably to final accuracy according to the ablation results.
  • The fused output improves upon existing single-instrument or simpler fusion retrieval baselines.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same staged architecture could be tested on data from other climate regimes to check whether the reported errors generalize.
  • The differentiable forward models open the possibility of joint training with additional sensor types not used in the original experiments.
  • The motion estimation step may enable short-term nowcasting applications if the reconstruction latency is reduced.
  • Extending the framework to include additional microphysical variables such as ice water content would test the limits of the refinement module.

Load-bearing premise

The conditional variational refinement module produces physically consistent microphysical fields when optimized under the differentiable radar and image forward models.

What would settle it

Independent validation measurements at the same or similar sites that show liquid water content mean absolute error above 0.026 g m^{-3} or wind speed error above 1.18 m s^{-1} would falsify the reported accuracy.

Figures

Figures reproduced from arXiv: 2606.30647 by Xinze Zhang.

Figure 1
Figure 1. Figure 1: Multi-sensor observation gap motivating this work. (a) Spatial versus temporal resolution of individual instruments; [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the AtmoFuseNet pipeline. Multi-view images and instrument observations enter the CMHFA stage, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Per-metric comparison of cloud property retrieval. Each group shows one evaluation metric; the rightmost bar in [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 6
Figure 6. Figure 6: Effect of progressively adding physics loss terms. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 5
Figure 5. Figure 5: Normalized performance of each incremental model [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Motion estimation ablation. (a) Wind speed MAE. (b) Wind direction MAE. (c) LWC MAE; the dashed line [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: LWC MAE as a function of the number of input [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: LWC MAE (g m−3 ) stratified by cloud thick￾ness (left three columns) and cloud base height (right three columns). Cell color scales from green (low error) to red (high error). 4.4 Analytical Experiments 4.4.1 Sensitivity to Number of Input Views Reducing the number of cameras degrades accuracy gradu￾ally rather than abruptly. With just two views, AtmoFuseNet still reaches 0.048 g m−3 LWC MAE, which is bett… view at source ↗
read the original abstract

Dense volumetric reconstruction of cloud microphysical fields from sparse ground-based instruments remains an open problem, largely because the available measurements are heterogeneous in both modality and spatial coverage. We present AtmoFuseNet, a framework that fuses multi-view sky camera imagery with millimeter-wave cloud radar and ceilometer observations to produce 4D (three spatial dimensions plus time) estimates of cloud state and wind. The method operates in three stages: a cross-modal hierarchical aggregation module that combines image feature pyramids with instrument-derived vertical profiles through layer-wise cross-attention; a conditional variational refinement module that maps the resulting volume to physically consistent microphysical fields under differentiable radar and image forward models; and a correlation-based motion estimator that recovers per-voxel 3D wind vectors from consecutive volumetric reconstructions. On collocated observations from a semi-arid site, AtmoFuseNet reaches 0.026 g m^-3 liquid water content MAE and 1.18 m s^-1 wind speed MAE, improving over existing retrieval baselines. Ablation experiments isolate the contribution of each module.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces AtmoFuseNet, a three-stage neural framework for dense 4D volumetric reconstruction of cloud microphysical fields (liquid water content, etc.) and 3D wind vectors from heterogeneous ground-based sensors: multi-view sky cameras, millimeter-wave cloud radar, and ceilometer. Stage 1 performs cross-modal hierarchical aggregation via layer-wise cross-attention between image feature pyramids and vertical profiles; Stage 2 applies conditional variational refinement optimized against differentiable radar and image forward models; Stage 3 estimates motion via correlation. On collocated observations from a semi-arid site the method reports MAEs of 0.026 g m^{-3} (LWC) and 1.18 m s^{-1} (wind speed), outperforming retrieval baselines, supported by ablation studies.

Significance. If the physical-consistency claims of the variational refinement stage can be substantiated with closure tests, the work would represent a meaningful advance in multi-modal atmospheric remote sensing by enabling physically constrained 4D cloud-state estimation from sparse ground instruments. The absence of such verification currently limits the assessed significance.

major comments (2)
  1. [Abstract] The claim that the conditional variational refinement produces physically consistent microphysical fields rests on optimization under differentiable forward models, yet no forward-model residual statistics, closure-test results, or quantitative checks on held-out collocated observations are reported. This verification is load-bearing for interpreting the reported MAEs as evidence of successful 4D reconstruction rather than statistical memorization.
  2. [Abstract] Numerical results are presented without accompanying information on dataset size, number of collocated samples, definition of the retrieval baselines, error bars, cross-validation procedure, or details of the differentiable forward models, preventing independent evaluation of the stated improvements (0.026 g m^{-3} LWC MAE, 1.18 m s^{-1} wind MAE).
minor comments (1)
  1. The manuscript title appears truncated ("Cross-Modal Hierarchical Fusion for from Multi-Sensor Ground Observation"); consider completing it for clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight important aspects of reproducibility and physical consistency. We will revise the manuscript to incorporate the requested details and verifications as outlined below.

read point-by-point responses

  1. Referee: [Abstract] The claim that the conditional variational refinement produces physically consistent microphysical fields rests on optimization under differentiable forward models, yet no forward-model residual statistics, closure-test results, or quantitative checks on held-out collocated observations are reported. This verification is load-bearing for interpreting the reported MAEs as evidence of successful 4D reconstruction rather than statistical memorization.

    Authors: We agree that the current version does not report forward-model residual statistics, closure-test results, or quantitative checks on held-out data. While the variational stage is optimized against the differentiable forward models, these explicit verifications are absent. In revision we will add a dedicated analysis section with residual statistics, closure tests on held-out collocated observations, and quantitative consistency checks to support the physical-consistency claims and strengthen interpretation of the reported MAEs. revision: yes

  2. Referee: [Abstract] Numerical results are presented without accompanying information on dataset size, number of collocated samples, definition of the retrieval baselines, error bars, cross-validation procedure, or details of the differentiable forward models, preventing independent evaluation of the stated improvements (0.026 g m^{-3} LWC MAE, 1.18 m s^{-1} wind MAE).

    Authors: We acknowledge that these experimental details are not summarized in the abstract and are insufficiently highlighted for independent evaluation. The full manuscript contains some of this information in the methods and experiments sections, but we will expand both the abstract and main text to report dataset size, number of collocated samples, explicit definitions of the retrieval baselines, error bars derived from cross-validation, the cross-validation procedure, and additional specifics of the differentiable forward models. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected; results presented as empirical outcomes.

full rationale

The provided abstract and context describe a three-stage pipeline with MAEs reported on collocated observations, treated as independent empirical results against baselines. No equations, fitting procedures, or self-citations are quoted that would reduce any prediction or claim to its inputs by construction. The conditional variational refinement is described as an optimization step under forward models, but without specific self-referential definitions or load-bearing self-citations that create circularity. This aligns with the default expectation that most papers lack circularity; the reported improvements are not shown to be forced by definition or prior self-work.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented physical entities; the method implicitly assumes standard neural-network training and the existence of differentiable forward models for radar and imagery.

pith-pipeline@v0.9.1-grok · 5703 in / 1113 out tokens · 25436 ms · 2026-07-01T07:17:02.567916+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

58 extracted references · 5 canonical work pages

  1. [1]

    M. C. Allmen and W. P. Kegelmeyer. The computa- tion of cloud-base height from paired whole-sky imag- ing cameras.Journal of Atmospheric and Oceanic Tech- nology, 13, 1996

    1996

  2. [2]

    Accurate medium-range global weather forecasting with 3D neural networks

    Kaifeng Bi, Lingxi Xie, Hengheng Zhang, Xin Chen, Xiaotao Gu, and Qi Tian. Accurate medium-range global weather forecasting with 3D neural networks. Nature, 619(7970):533–538, 2023

    2023

  3. [3]

    MVS- NeRF: Fast generalizable radiance field reconstruc- tion from multi-view stereo

    Anpei Chen, Zexiang Xu, Fuqiang Zhao, Xiaosong Zhang, Fanbo Xiang, Jingyi Yu, and Hao Su. MVS- NeRF: Fast generalizable radiance field reconstruc- tion from multi-view stereo. InProceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 14124–14133, 2021

    2021

  4. [4]

    Intent: Invari- ance and discrimination-aware noise mitigation for ro- bust composed image retrieval

    Zhiwei Chen, Yupeng Hu, Zhiheng Fu, Zixu Li, Jiale Huang, Qinlei Huang, and Yinwei Wei. Intent: Invari- ance and discrimination-aware noise mitigation for ro- bust composed image retrieval. InProceedings of the AAAI Conference on Artificial Intelligence, 2026

    2026

  5. [5]

    Invariance and discrimination-aware noise mitigation for robust com- posed image retrieval

    Zhiwei Chen, Yupeng Hu, Zhiheng Fu, Zixu Li, Jiale Huang, Qinlei Huang, and Yinwei Wei. Invariance and discrimination-aware noise mitigation for robust com- posed image retrieval. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 20463–20471, 2026

    2026

  6. [6]

    Learning phrase rep- resentations using RNN encoder–decoder for statistical machine translation

    Kyunghyun Cho, Bart van Merrienboer, Caglar Gul- cehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. Learning phrase rep- resentations using RNN encoder–decoder for statistical machine translation. InProceedings of the 2014 Confer- ence on Empirical Methods in Natural Language Pro- cessing, 2014

    2014

  7. [7]

    4D spatio-temporal ConvNets: Minkowski convolutional neural networks

    Christopher Choy, JunYoung Gwak, and Silvio Savarese. 4D spatio-temporal ConvNets: Minkowski convolutional neural networks. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3075–3084, 2019

    2019

  8. [8]

    Savoy, Yee Hui Lee, and Stefan Winkler

    Soumyabrata Dev, Florian M. Savoy, Yee Hui Lee, and Stefan Winkler. CloudSegNet: A deep network for ny- chthemeron cloud image segmentation.IEEE Transac- tions on Geoscience and Remote Sensing, 57(12):9410– 9422, 2019

    2019

  9. [9]

    FlowNet: Learning optical flow with convolutional net- works

    Alexey Dosovitskiy, Philipp Fischer, Eddy Ilg, Philip Hausser, Caner Hazirbas, Vladimir Golkov, Patrick 9 van der Smagt, Daniel Cremers, and Thomas Brox. FlowNet: Learning optical flow with convolutional net- works. InProceedings of the IEEE International Con- ference on Computer Vision, pages 2758–2766, 2015

    2015

  10. [10]

    Academic Press, 2nd edi- tion, 1993

    Richard J Doviak and Du ˇsan S Zrni ´c.Doppler Radar and Weather Observations. Academic Press, 2nd edi- tion, 1993

    1993

  11. [11]

    Scott Frisch, Graham Feingold, Christopher W

    A. Scott Frisch, Graham Feingold, Christopher W. Fairall, Taneil Uttal, and Jefferson B. Snider. On cloud radar and microwave radiometer measurements of stra- tus cloud liquid water profiles.Journal of Geophysical Research: Atmospheres, 103(D19):23195–23197, 1998

    1998

  12. [12]

    Air-know: Arbiter-calibrated knowledge-internalizing robust network for composed image retrieval

    Zhiheng Fu, Yupeng Hu, Qianyun Yang, Shiqi Zhang, Zhiwei Chen, and Zixu Li. Air-know: Arbiter-calibrated knowledge-internalizing robust network for composed image retrieval. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR), 2026

    2026

  13. [13]

    Modeling the dynamics of PDE systems with physics-constrained deep auto-regressive networks.Journal of Computa- tional Physics, 403:109056, 2020

    Nicholas Geneva and Nicholas Zabaras. Modeling the dynamics of PDE systems with physics-constrained deep auto-regressive networks.Journal of Computa- tional Physics, 403:109056, 2020

    2020

  14. [14]

    Cambridge University Press, 2nd edition, 2003

    Richard Hartley and Andrew Zisserman.Multiple View Geometry in Computer Vision. Cambridge University Press, 2nd edition, 2003

    2003

  15. [15]

    Au- tomatic cloud classification of whole sky images.Atmo- spheric Measurement Techniques, 3(3):557–567, 2010

    Anna Heinle, Andreas Macke, and Anand Srivastav. Au- tomatic cloud classification of whole sky images.Atmo- spheric Measurement Techniques, 3(3):557–567, 2010

    2010

  16. [16]

    The ERA5 global reanalysis.Quarterly Journal of the Royal Meteorological Society, 146(730):1999–2049, 2020

    Hans Hersbach, Bill Bell, Paul Berrisford, Shoji Hira- hara, Andr ´as Hor ´anyi, Joaqu ´ın Mu˜noz-Sabater, Julien Nicolas, Carole Peubey, Raluca Radu, Dinand Schepers, et al. The ERA5 global reanalysis.Quarterly Journal of the Royal Meteorological Society, 146(730):1999–2049, 2020

    1999

  17. [17]

    Denois- ing diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denois- ing diffusion probabilistic models. InAdvances in Neu- ral Information Processing Systems, volume 33, pages 6840–6851, 2020

    2020

  18. [18]

    Horn and Brian G

    Berthold K.P. Horn and Brian G. Schunck. Determining optical flow.Artificial Intelligence, 17(1–3):185–203, 1981

    1981

  19. [19]

    Refine: Com- posed video retrieval via shared and differential seman- tics enhancement.ACM Transactions on Multimedia Computing, Communications and Applications, 2026

    Yupeng Hu, Zixu Li, Zhiwei Chen, Qinlei Huang, Zhi- heng Fu, Mingzhu Xu, and Liqiang Nie. Refine: Com- posed video retrieval via shared and differential seman- tics enhancement.ACM Transactions on Multimedia Computing, Communications and Applications, 2026

    2026

  20. [20]

    Kevrekidis, Lu Lu, Paris Perdikaris, Sifan Wang, and Liu Yang

    George Em Karniadakis, Ioannis G. Kevrekidis, Lu Lu, Paris Perdikaris, Sifan Wang, and Liu Yang. Physics- informed machine learning.Nature Reviews Physics, 3(6):422–440, 2021

    2021

  21. [21]

    Kingma and Max Welling

    Diederik P. Kingma and Max Welling. Auto-encoding variational bayes. InInternational Conference on Learning Representations, 2014

    2014

  22. [22]

    Learning skillful medium-range global weather forecasting.Science, 382(6677):1416–1421, 2023

    Remi Lam, Alvaro Sanchez-Gonzalez, Matthew Will- son, Peter Wirnsberger, Meire Fortunato, Ferran Alet, Suman Ravuri, Timo Ewalds, Zach Eaton-Rosen, Wei- hua Hu, et al. Learning skillful medium-range global weather forecasting.Science, 382(6677):1416–1421, 2023

    2023

  23. [23]

    Airborne three-dimensional cloud to- mography

    Aviad Levis, Yoav Y Schechner, Amit Aides, and An- thony B Davis. Airborne three-dimensional cloud to- mography. InProceedings of the IEEE International Conference on Computer Vision, pages 3379–3387, 2015

    2015

  24. [24]

    Multiple-scattering microphysics tomography

    Aviad Levis, Yoav Y Schechner, and Anthony B Davis. Multiple-scattering microphysics tomography. InPro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 6740–6749, 2017

    2017

  25. [25]

    Multi-view polarimetric scattering cloud tomography and retrieval of droplet size

    Aviad Levis, Yoav Y Schechner, and Anthony B Davis. Multi-view polarimetric scattering cloud tomography and retrieval of droplet size. InEuropean Conference on Computer Vision, 2020

    2020

  26. [26]

    Exploring efficient open-vocabulary segmentation in the remote sensing,

    Bingyu Li, Haocheng Dong, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li. Exploring efficient open- vocabulary segmentation in the remote sensing.arXiv preprint arXiv:2509.12040, 2025

    work page arXiv 2025

  27. [27]

    Exploring the underwater world segmentation without extra training,

    Bingyu Li, Tao Huo, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li. Exploring the underwater world segmentation without extra training.arXiv preprint arXiv:2511.07923, 2025

    work page arXiv 2025

  28. [28]

    Maris: Marine open-vocabulary instance segmentation with geometric enhancement and semantic alignment,

    Bingyu Li, Feiyu Wang, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li. Maris: Marine open- vocabulary instance segmentation with geometric en- hancement and semantic alignment.arXiv preprint arXiv:2510.15398, 2025

    work page arXiv 2025

  29. [29]

    Stitchfusion: Weaving any visual modal- ities to enhance multimodal semantic segmentation

    Bingyu Li, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li. Stitchfusion: Weaving any visual modal- ities to enhance multimodal semantic segmentation. In Proceedings of the 33rd ACM International Conference on Multimedia, pages 1308–1317, 2025

    2025

  30. [30]

    U3m: Unbiased multiscale modal fusion model for multimodal semantic segmentation.Pattern Recognition, 168:111801, 2025

    Bingyu Li, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li. U3m: Unbiased multiscale modal fusion model for multimodal semantic segmentation.Pattern Recognition, 168:111801, 2025

    2025

  31. [31]

    Retrack: Evidence-driven dual-stream directional anchor calibra- tion network for composed video retrieval

    Zixu Li, Yupeng Hu, Zhiwei Chen, Qinlei Huang, Guozhi Qiu, Zhiheng Fu, and Meng Liu. Retrack: Evidence-driven dual-stream directional anchor calibra- tion network for composed video retrieval. InProceed- ings of the AAAI Conference on Artificial Intelligence, 2026. 10

    2026

  32. [32]

    Conesep: Cone-based ro- bust noise-unlearning compositional network for com- posed image retrieval

    Zixu Li, Yupeng Hu, Zhiwei Chen, Mingyu Zhang, Zhi- heng Fu, and Liqiang Nie. Conesep: Cone-based ro- bust noise-unlearning compositional network for com- posed image retrieval. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR), 2026

    2026

  33. [33]

    Habit: Chrono-synergia robust progressive learning framework for composed image retrieval

    Zixu Li, Yupeng Hu, Zhiwei Chen, Shiqi Zhang, Qin- lei Huang, Zhiheng Fu, and Yinwei Wei. Habit: Chrono-synergia robust progressive learning framework for composed image retrieval. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 6762–6770, 2026

    2026

  34. [34]

    Habit: Chrono-synergia robust progressive learning framework for composed image retrieval

    Zixu Li, Yupeng Hu, Zhiwei Chen, Shiqi Zhang, Qin- lei Huang, Zhiheng Fu, and Yinwei Wei. Habit: Chrono-synergia robust progressive learning framework for composed image retrieval. InProceedings of the AAAI Conference on Artificial Intelligence, 2026

    2026

  35. [35]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. InInternational Conference on Learning Representations, 2019

    2019

  36. [36]

    NeRF: Representing scenes as neural radiance fields for view synthesis

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. NeRF: Representing scenes as neural radiance fields for view synthesis. InEuropean Conference on Computer Vision, pages 405–421. Springer, 2020

    2020

  37. [37]

    Determina- tion of the optical thickness and effective particle radius of clouds from reflected solar radiation measurements

    Teruyuki Nakajima and Michael D King. Determina- tion of the optical thickness and effective particle radius of clouds from reflected solar radiation measurements. Part I: Theory.Journal of the Atmospheric Sciences, 47(15):1878–1893, 1990

    1990

  38. [38]

    King, Steven A

    Steven Platnick, Michael D. King, Steven A. Ackerman, W. Paul Menzel, Bryan A. Baum, J ´erˆome C. Ri´edi, and Rong A. Frey. The MODIS cloud products: Algorithms and examples from terra.IEEE Transactions on Geo- science and Remote Sensing, 41(2):459–473, 2003

    2003

  39. [39]

    Andersson, Andrew El-Kadi, Dominic Masters, Timo Ewalds, Jacklynn Stott, Shakir Mohamed, Peter Battaglia, Remi Lam, Matthew Willson, et al

    Ilan Price, Alvaro Sanchez-Gonzalez, Ferran Alet, Tom R. Andersson, Andrew El-Kadi, Dominic Masters, Timo Ewalds, Jacklynn Stott, Shakir Mohamed, Peter Battaglia, Remi Lam, Matthew Willson, et al. Proba- bilistic weather forecasting with machine learning.Na- ture, 637(8044):84–90, 2025

    2025

  40. [40]

    Maziar Raissi, Paris Perdikaris, and George E Karni- adakis. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equa- tions.Journal of Computational Physics, 378:686–707, 2019

    2019

  41. [41]

    VIP-CT: Variable importance parametric cloud tomog- raphy

    Roi Ronen, Vadim Holodovsky, and Yoav Y Schechner. VIP-CT: Variable importance parametric cloud tomog- raphy. InEuropean Conference on Computer Vision, 2022

    2022

  42. [42]

    U-Net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-Net: Convolutional networks for biomedical image segmentation. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 234–241. Springer, 2015

    2015

  43. [43]

    Reasoning in computer vision: Taxonomy, models, tasks, and methodologies.arXiv preprint arXiv:2508.10523, 2025

    Ayushman Sarkar, Mohd Yamani Idna Idris, and Zhenyu Yu. Reasoning in computer vision: Taxonomy, models, tasks, and methodologies.arXiv preprint arXiv:2508.10523, 2025

    work page arXiv 2025

  44. [44]

    A large eddy simula- tion intercomparison study of shallow cumulus convec- tion.Journal of the Atmospheric Sciences, 60(10):1201– 1219, 2003

    A Pier Siebesma, Christopher S Bretherton, An- drew Brown, Andreas Chlond, Joan Cuxart, Peter G Duynkerke, Hongli Jiang, Marat Khairoutdinov, David Lewellen, Chin-Hoh Moeng, et al. A large eddy simula- tion intercomparison study of shallow cumulus convec- tion.Journal of the Atmospheric Sciences, 60(10):1201– 1219, 2003

    2003

  45. [45]

    Learning structured output representation using deep conditional generative models

    Kihyuk Sohn, Honglak Lee, and Xinchen Yan. Learning structured output representation using deep conditional generative models. InAdvances in Neural Information Processing Systems, 2015

    2015

  46. [46]

    De- noising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. De- noising diffusion implicit models. InInternational Con- ference on Learning Representations, 2021

    2021

  47. [47]

    Stephens

    Graeme L. Stephens. Cloud feedbacks in the cli- mate system: A critical review.Journal of Climate, 18(2):237–273, 2005

    2005

  48. [48]

    Mingxing Tan and Quoc V . Le. EfficientNet: Rethink- ing model scaling for convolutional neural networks. In International Conference on Machine Learning. PMLR, 2019

    2019

  49. [49]

    RAFT: Recurrent all-pairs field transforms for optical flow

    Zachary Teed and Jia Deng. RAFT: Recurrent all-pairs field transforms for optical flow. InEuropean Con- ference on Computer Vision, pages 402–419. Springer, 2020

    2020

  50. [50]

    van Heerwaarden, Bart J

    Chiel C. van Heerwaarden, Bart J. H. van Stratum, Thijs Heus, Jeremy A. Gibbs, Evgeni Fedorovich, and Juan Pedro Mellado. MicroHH 1.0: A computational fluid dynamics code for direct numerical simulation and large-eddy simulation of atmospheric boundary layer flows.Geoscientific Model Development, 10(8):3145– 3165, 2017

    2017

  51. [51]

    Three-dimensional scene flow

    Sundar Vedula, Simon Baker, Peter Rander, Robert Collins, and Takeo Kanade. Three-dimensional scene flow. InIEEE International Conference on Computer Vision, 1999

    1999

  52. [52]

    Velden, Christopher M

    Christopher S. Velden, Christopher M. Hayden, Steven J. Nieman, W. Paul Menzel, Steve Wanzong, and James S. Goerss. Recent innovations in deriving tro- pospheric winds from meteorological satellites.Bulletin of the American Meteorological Society, 86(2):205–223, 2005. 11

    2005

  53. [53]

    MVSNet: Depth inference for unstructured multi- view stereo

    Yao Yao, Zixin Luo, Shiwei Li, Tian Fang, and Long Quan. MVSNet: Depth inference for unstructured multi- view stereo. InProceedings of the European Conference on Computer Vision, pages 767–783, 2018

    2018

  54. [54]

    pixelNeRF: Neural radiance fields from one or few images

    Alex Yu, Vickie Ye, Matthew Tancik, and Angjoo Kanazawa. pixelNeRF: Neural radiance fields from one or few images. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021

    2021

  55. [55]

    From physics to foundation models: A review of ai-driven quan- titative remote sensing inversion.arXiv preprint arXiv:2507.09081, 2025

    Zhenyu Yu, Mohd Yamani Idna Idris, Hua Wang, Pei Wang, Junyi Chen, and Kun Wang. From physics to foundation models: A review of ai-driven quan- titative remote sensing inversion.arXiv preprint arXiv:2507.09081, 2025

    work page arXiv 2025

  56. [56]

    Dinov3-powered multi-task founda- tion model for quantitative remote sensing estimation (student abstract)

    Zhenyu Yu, Mohd Yamani Idna Idris, Pei Wang, and Rizwan Qureshi. Dinov3-powered multi-task founda- tion model for quantitative remote sensing estimation (student abstract). InProceedings of the AAAI Confer- ence on Artificial Intelligence, volume 40, pages 41455– 41456, 2026

    2026

  57. [57]

    Spatiotemporal alignment for remote sens- ing image recovery via terrain-aware diffusion

    Zhenyu Yu, Haoran Jiang, Pei Wang, Zizhen Lin, and Yong Xiang. Spatiotemporal alignment for remote sens- ing image recovery via terrain-aware diffusion. In ICASSP 2026-2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 11257–11261. IEEE, 2026

    2026

  58. [58]

    Qrs-trs: Style transfer-based image- to-image translation for carbon stock estimation in quan- titative remote sensing.IEEE Access, 2025

    Zhenyu Yu, Jinnian Wang, Hanqing Chen, and Mohd Yamani Idna Idris. Qrs-trs: Style transfer-based image- to-image translation for carbon stock estimation in quan- titative remote sensing.IEEE Access, 2025. 12

    2025