Pith. sign in

REVIEW 4 major objections 5 minor 27 references

Cloud Optical Thickness Retrievals Using Angle Invariant Attention Based Deep Learning Models

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read CAAC, an attention-based neural network with angle embeddings, is claimed to retrieve cloud optical thickness from two-wavelength radiance with at least ninefold lower mean squared error than single-angle baselines.

desk verdict A sensible angle-conditioned U-Net for cloud optical thickness retrieval, but the 9x headline overstates Table 2 and the missing train/test split description makes the main result hard to trust. read the letter →

arxiv 2505.24638 v1 pith:QP6S75YE submitted 2025-05-30 cs.CV cs.AI

Zahid Hassan Tushar , Adeleke Ademakinwa , Jianwu Wang , Zhibo Zhang , Sanjay Purushotham This is my paper
classification cs.CVcs.AI
keywords cloudopticalthicknessdeeplearningretrievalattentionmechanismangleinvariancemulti-angletrainingremotesensingradiativetransferUNet
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to show that cloud optical thickness (COT), the quantity that controls how much sunlight a cloud reflects, can be retrieved from two-wavelength satellite radiance without losing accuracy as the sun's angle or the satellite's viewing angle changes. The proposed model, Cloud-Attention-Net with Angle Coding (CAAC), is a compact UNet-style network whose attention gating sharpens cloud features and whose angle embeddings are injected through adaptive instance normalization. On a simulated dataset built from large-eddy-simulation cloud fields and a 3D radiative transfer model, the paper reports that multi-angle trained CAAC has the lowest mean squared error (MSE) of all methods tested, with a ninefold or larger reduction in MSE reported against the Independent Pixel Approximation and the single-angle deep learning baselines. The paper also reports that CAAC remains accurate on combinations of sun and viewing angles it never saw in training, which is the condition a real satellite retrieval would face. A reader should care because cloud optical thickness retrieval errors feed directly into biases in climate and weather model evaluations, and current deep learning baselines either ignore viewing geometry or must be retrained for each angle.

What carries the argument

The central object is the CAAC architecture: a UNet whose skip connections pass through a Channel Attention Module (CAM) that reweights feature maps channel by channel, and whose decoder layers receive the viewing geometry through Adaptive Instance Normalization (AdaIN). The AdaIN operation is the identity $$\mathrm{AdaIN}(x,y)=\$\sigma$(y)\,((x-\mu(x))/\$\sigma$(x))+\mu(y)$$, where $x$ is a feature map and $y$ is the angle information: the feature map is normalized by its own mean and standard deviation, then rescaled and shifted by statistics learned from the angle. This one operation is what lets a single shared network adapt its features to different solar and viewing angles instead of needing one model per angle. Multi-angle training supplies the angle-conditioned statistics, while the channel attention is what the paper credits with preserving informative cloud-structure features along the encoder-decoder path.

What would settle it

Train CAAC and all baselines on 64-by-64 windows from some LES cloud fields, hold out complete cloud fields for testing, and recompute the MSE ratios. If the ninefold advantage over single-angle methods shrinks substantially, or if CAAC no longer beats multi-angle trained UNet, the central claim is contradicted.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the same cloud field viewed under different solar zenith and view zenith angles produces radiance fields with changed intensity, shadowing, and apparent horizontal shifts, and that a retrieval model can be made invariant to those changes by conditioning its internal features on the angles themselves. The paper claims that CAAC, trained jointly on four solar zenith angles and three view zenith angles, outperforms the Independent Pixel Approximation, single-angle trained UNet and CloudUNet, and multi-angle trained versions of those networks on every data partition reported. It gives the headline margins as ninefold, sevenfold, and elevenfold lower MSE than the single-angle methods, and attributes the gain to channel attention in place of plain skip connections plus AdaIN-based angle coding. In a separate experiment on cloud profiles and angles not used in training, CAAC is reported to give the best and most stable performance across the tested grid of sun and viewing angles. The evidence is synthetic but physically motivated: large-eddy-simulation cloud fields with radiance computed by a 3D radiative transfer solver at 0.66 and 2.13 microns.

Load-bearing premise

The load-bearing premise is that the reported error reductions are not inflated by the unspecified train/test split of the overlapping 64-by-64 image windows, since the paper does not say whether entire cloud fields were held out from training.

Editorial extensions

If this is right

  • One model trained on multiple angles could replace angle-specific lookup tables or per-angle trained networks, saving memory and compute in operational satellite pipelines.
  • Angle-conditioned retrieval should reduce COT bias at large solar and viewing angles, exactly where the Independent Pixel Approximation shows the most over- and under-estimation due to shadows and cloud distortion.
  • The same architecture can be extended to jointly retrieve cloud optical thickness and effective radius in multi-angle settings, which is listed as the immediate next step.
  • If the reported generalization to unseen angles holds, satellite instruments with varying geometry could be processed without retraining for every possible solar and view configuration.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The ninefold headline is anchored to single-angle baselines; the paper's own table shows a smaller margin over multi-angle trained UNet and CloudUNet, so the practical size of the gain depends on which baseline is the point of comparison.
  • Because the paper does not specify whether the 64-by-64 windows were split by cloud field or randomly, and the windows overlap with stride 10, the reported factors should be treated as upper bounds until a whole-cloud-field held-out test confirms them.
  • The angle-conditioning mechanism is not specific to cloud optics: the same AdaIN-plus-attention pattern could be tested on other remote-sensing inversions whose signal is distorted by viewing geometry, such as aerosol optical depth or surface reflectance.
  • Demonstrating the model on real satellite radiance with independent cloud references, rather than simulated LES radiance, is the natural next test of whether the angle invariance survives the gap between simulated and observed scenes.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes CAAC, a UNet-style deep learning model with channel attention and AdaIN-based angle conditioning, for retrieving cloud optical thickness (COT) from two-wavelength radiance images. Experiments use 102 LES cloud fields with SHDOM-simulated radiances at multiple solar and view zenith angles, evaluated against IPA, UNet, and CloudUNet baselines. The central reported result is that multi-angle CAAC achieves 0.0459 MSE on the all-angle partition, which the authors describe as a nine-fold improvement over single-angle methods and IPA; the abstract and conclusion generalize this to 'at least a factor of nine' reduction in retrieval errors.

Significance. If the claims hold, the work is a useful step in angle-robust COT retrieval: the use of 3D radiative transfer simulation, explicit angle conditioning, and multi-angle training is sensible, and the ablations (CAAC with/without attention and angle coding) help isolate the contribution of each component. The paper's strengths include the physically grounded simulated dataset, the clear architectural motivation, and the attempt to evaluate on unseen angles. However, the headline factor-of-nine claim is overstated relative to the multi-angle baselines in Table 2, and the paper omits a crucial train/test split description, which currently prevents the main quantitative result from being independently verified.

major comments (4)
  1. [Section 4, Implementation details] The paper does not state whether the train/validation/test split is performed at the cloud-field level or at the level of individual windows. With 102 LES cloud fields and 64x64 windows extracted at stride 10, each field yields roughly 64 overlapping windows; a random window split would place spatially overlapping windows from the same field in both training and test sets, allowing the model to memorize local spatial structure and inflating the reported MSE improvements. Please specify the split granularity, report the number of test fields and test windows, and provide results under a field-level split or repeated cross-validation.
  2. [Section 5, Table 2 and abstract] The 'at least a factor of nine' claim is not supported by the full comparison in Table 2. In the SZA:all, VZA:all partition, CAAC (0.0459 MSE) is only 1.48x lower than multi-angle UNet (0.0678) and 1.69x lower than multi-angle CloudUNet (0.0778); the 9x figure arises only when comparing against IPA (0.4212) and single-angle UNet/CloudUNet (0.4097). The abstract, Section 5, and conclusion should state the baseline-specific ratios rather than a blanket 'at least nine-fold' improvement.
  3. [Section 5, Evaluation on unseen SZA and VZA] The angle-invariance claim rests primarily on Figure 7, which is presented only as a heatmap with no numerical MSE values or comparison table for the 165 unseen angle combinations. It is also not stated whether the 'different set of LES cloud profiles' used in this evaluation is disjoint from the profiles used during training. Please provide quantitative results and confirm field-level disjointness.
  4. [Section 3 and Figure 2] The angle-coding mechanism is not fully specified. AdaIN is defined in Eq. (1), but the paper does not describe how the scalar SZA and VZA values are embedded, what network maps the angle code to the AdaIN mean and standard deviation, or at which decoder resolutions the AdaIN blocks are applied. Without this information, CAAC is not reproducible from the manuscript.
minor comments (5)
  1. [Title and Table 1] There are typographical errors: the title contains 'RETRIEV ALS' instead of 'RETRIEVALS', and Table 1 has the header 'Parititon' instead of 'Partition'.
  2. [Section 4, Data Pre-processing] The shifted log transformation is written as log(COT+1); please specify whether this is the natural logarithm and clarify whether the reported MSE is computed on the transformed scale or after inverse transformation.
  3. [Figure 3] Figure 3 lacks axis labels and a precise explanation of what '20% of multi-angle training data' means, which makes the reported 3.4x and 7.5x improvements difficult to interpret or reproduce.
  4. [References] Reference [7] lists 'Rintaro Okamura et al.' without the full author list; please complete the citation.
  5. [General] The paper would benefit from a data/code availability statement and from explicit reporting of the number of training, validation, and test windows used in Table 2.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: CAAC retrieval is supervised regression against independent simulated COT ground truth, and the 9x claim is an arithmetic ratio from Table 2.

full rationale

The paper's central claim is that the CAAC model achieves lower MSE than IPA, UNet, and CloudUNet on simulated LES/SHDOM radiance data. The target COT values are produced by the LASSO LES cloud fields and SHDOM radiative transfer, not by the model or by the fitting procedure. No equation defines the predicted COT in terms of the ground-truth COT, and no fitted parameter is renamed as a prediction. The '9x, 7x, 11x lower MSE' statement in Section 5 is an arithmetic ratio of the Table 2 entries (e.g., 0.4212/0.0459 for IPA versus CAAC), so it is a reported comparison, not a derived prediction. The unseen-angle experiment on SZAs [5,10,20,30,50] and 33 VZAs is a genuine generalization test rather than a re-fit, since those angles were not in the training set. Self-citations appear ([9] CloudUNet is used as the window framework and as a baseline; [2] is prior work on 3D radiative effects), but they do not supply the retrieval result; the baseline comparison is against independent SHDOM ground truth. One reproducibility caveat is that Section 4 does not state whether 64x64 stride-10 windows were split at cloud-field level or randomly, which could affect the magnitude of the reported gains; this is a data-handling concern, not circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

No new physical entities are introduced. The central claim rests on the representativeness of simulated LES/SHDOM data and on hand-chosen architectural hyperparameters; no physics-derived constants are fitted to the retrieval target.

free parameters (4)
  • Window size = 64x64
    Chosen by hand as a common practice in window-based retrieval; affects spatial context and number of training samples.
  • Stride = 10
    Chosen by hand; controls the number of extracted windows and creates overlapping patches used as training and evaluation samples.
  • Architecture channel widths = 128, 64, 32
    Model capacity selected by hand; influences accuracy and computational cost.
  • Learning rate = searched in [1e-3, 1e-2, 1e-1]
    Tuning range chosen by hand; final learning rate is selected by validation performance.
assumptions (5)
  • domain assumption SHDOM simulations at 0.66 and 2.13 micrometers accurately represent radiance from realistic 3D clouds.
    Section 2 uses SHDOM as the sole source of radiance observations for training and evaluation; no real satellite radiance is used.
  • domain assumption LASSO LES cloud fields with a constant effective radius of 12 micrometers are representative of real clouds for retrieval.
    Section 2 derives all 102 cloud profiles from LASSO LES with fixed droplet effective radius, ignoring variability in droplet size distribution.
  • domain assumption The surface behaves as a Lambertian reflector with albedo 0.05 in both spectral channels.
    Set in Section 2 for all radiative transfer calculations; real land and ocean surfaces are not Lambertian at 0.66 and 2.13 micrometers.
  • domain assumption Two spectral bands, 0.66 and 2.13 micrometers, provide sufficient information to retrieve cloud optical thickness.
    The regression input is limited to these two bands, following the window-based retrieval framework from prior work; no sensitivity analysis for other bands is given.
  • domain assumption Ground truth COT from LES cloud fields is noise-free and perfectly aligned with the simulated radiance pixels.
    The model is trained to predict COT directly from SHDOM radiance at the same grid resolution; the paper does not discuss geolocation errors or instrument noise.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cloud Optical Thickness Retrievals Using Angle Invariant Attention Based Deep Learning Models." pith.science (2026). https://pith.science/paper/QP6S75YE

@misc{pith2026250524638,
  author       = {Pith},
  title        = {Pith review of: Cloud Optical Thickness Retrievals Using Angle Invariant Attention Based Deep Learning Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QP6S75YE}},
  note         = {Machine review of arXiv:2505.24638}
}
read the original abstract

Cloud Optical Thickness (COT) is a critical cloud property influencing Earth's climate, weather, and radiation budget. Satellite radiance measurements enable global COT retrieval, but challenges like 3D cloud effects, viewing angles, and atmospheric interference must be addressed to ensure accurate estimation. Traditionally, the Independent Pixel Approximation (IPA) method, which treats individual pixels independently, has been used for COT estimation. However, IPA introduces significant bias due to its simplified assumptions. Recently, deep learning-based models have shown improved performance over IPA but lack robustness, as they are sensitive to variations in radiance intensity, distortions, and cloud shadows. These models also introduce substantial errors in COT estimation under different solar and viewing zenith angles. To address these challenges, we propose a novel angle-invariant, attention-based deep model called Cloud-Attention-Net with Angle Coding (CAAC). Our model leverages attention mechanisms and angle embeddings to account for satellite viewing geometry and 3D radiative transfer effects, enabling more accurate retrieval of COT. Additionally, our multi-angle training strategy ensures angle invariance. Through comprehensive experiments, we demonstrate that CAAC significantly outperforms existing state-of-the-art deep learning models, reducing cloud property retrieval errors by at least a factor of nine.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

27 extracted references · 26 canonical work pages

  1. [1]

    Cloud Optical Thickness Retrievals Using Angle Invariant Attention Based Deep Learning Models

    INTRODUCTION Clouds play a crucial role in regulating the Earth’s radiation budget and are characterized by various microphysical prop- erties, including cloud optical thickness (COT), cloud effec- tive radius (CER), and cloud top height (CTH). These prop- erties are essential for understanding climate change and im- proving weather forecasting. Given the...

    work page Pith review arXiv 2025

  2. [2]

    Dataset: Satellites capture cloud radiance observations from real clouds at specific solar zenith angles (SZAs) and view zenith angles (VZAs)

    DATA AND PROBLEM FORMULATION In this section, we describe our dataset and the challenges associated with cloud property retrievals due to solar and view zenith angles, as well as our problem formulation. Dataset: Satellites capture cloud radiance observations from real clouds at specific solar zenith angles (SZAs) and view zenith angles (VZAs). However, a...

    work page

  3. [3]

    CLOUD-ATTENTION-NET WITH ANGLE CODING (CAAC) Recent advancements in cloud property retrieval algorithms from radiance observations have increasingly leveraged ma- chine learning and deep learning techniques. Among these, UNet-style architectures such as CloudUNet [9] have gained prominence due to their ability to extract 2D spatial features from radiance ...

    work page

  4. [4]

    EXPERIMENTS We conducted a comprehensive set of experiments to address the following research questions: (1) Model Comparison : How does our proposed CAAC model perform compared to existing COT retrieval methods? (2) Impact of SZA and VZA: What is the impact of SZA (Solar Zenith Angle) and VZA (View Zenith Angle) on the performance of COT retrieval models...

    work page

  5. [5]

    The results indicate that multi-angle training enhances the performance of all COT retrieval methods compared to single-angle training

    RESULTS AND DISCUSSION Table 2 shows the model performance (MSE scores) for all the COT retrieval methods across different data parti- tions. The results indicate that multi-angle training enhances the performance of all COT retrieval methods compared to single-angle training. Furthermore, our multi-angle CAAC model achieves the best overall performance, ...

    work page

  6. [6]

    CONCLUSION In this work, we introduced Cloud-Attention-Net with Angle Coding (CAAC)—the first angle-invariant, attention-based deep learning model for COT retrieval. We showed that our CAAC model which comprises of a compact U-Net architec- ture with attention and angle-coding modules, can effectively leverage angle information from radiance observations ...

    work page

  7. [7]

    National Academies of Sciences, Division on Engineering, Physical Sciences, Space Studies Board, Committee on the Decadal Survey for Earth Science, and Applications from Space, Thriving on our changing planet: A decadal strategy for Earth observation from space, National Academies Press, 2019

    work page 2019

  8. [8]

    Influence of cloud retrieval errors due to three dimensional radiative effects on calculations of broadband cloud radiative effect,

    Adeleke S Ademakinwa, Zahid H Tushar, Jianyu Zheng, Chenxi Wang, Sanjay Purushotham, Jianwu Wang, Kerry G Meyer, Tamas V´arnai, and Zhibo Zhang, “Influence of cloud retrieval errors due to three dimensional radiative effects on calculations of broadband cloud radiative effect,” EGUsphere, vol. 2023, pp. 1–32, 2023

    work page 2023

Show all 27 references
  1. [9]

    Determination of the optical thickness and effective particle radius of clouds from reflected solar radiation measurements. part i: Theory,

    Teruyuki Nakajima and Michael D King, “Determination of the optical thickness and effective particle radius of clouds from reflected solar radiation measurements. part i: Theory,” Journal of Atmospheric Sciences , vol. 47, no. 15, pp. 1878– 1893, 1990

    1990

  2. [10]

    Effect of scattering angle on earth reflectance,

    Alexander Marshak, Alfonso Delgado-Bonal, and Yuri Knyazikhin, “Effect of scattering angle on earth reflectance,” Frontiers in Remote Sensing, vol. 2, pp. 719610, 2021

    2021

  3. [11]

    Spectral anisotropy of subtropical deciduous forest using misr and modis data ac- quired under large seasonal variation in solar zenith angle,

    Fabio Marcelo Breunig, L ˆenio Soares Galv ˜ao, Jo ˜ao Roberto Dos Santos, Anatoly A Gitelson, Yhasmin Mendes De Moura, Thiago Sousa Teles, and William Gaida, “Spectral anisotropy of subtropical deciduous forest using misr and modis data ac- quired under large seasonal variati...

    2015

  4. [12]

    Effect of cloud inhomogeneities on the solar zenith angle dependence of nadir reflectance,

    Norman G Loeb, Tam ´as V´arnai, and Roger Davies, “Effect of cloud inhomogeneities on the solar zenith angle dependence of nadir reflectance,” Journal of Geophysical Research: Atmo- spheres, vol. 102, no. D8, pp. 9387–9395, 1997

    1997

  5. [13]

    Feasibility study of multi-pixel re- trieval of optical thickness and droplet effective radius of in- homogeneous clouds using deep learning,

    Rintaro Okamura et al., “Feasibility study of multi-pixel re- trieval of optical thickness and droplet effective radius of in- homogeneous clouds using deep learning,” Atmospheric Mea- surement Techniques, 2017

    2017

  6. [14]

    Segmentation-based multi-pixel cloud op- tical thickness retrieval using a convolutional neural network,

    Vikas Nataraja, Sebastian Schmidt, Hong Chen, Takanobu Ya- maguchi, Jan Kazil, Graham Feingold, Kevin Wolf, and Hi- ronobu Iwabuchi, “Segmentation-based multi-pixel cloud op- tical thickness retrieval using a convolutional neural network,” Atmospheric Measurement Techniques Di...

    2022

  7. [15]

    Cloudunet: Adapt- ing unet for retrieving cloud properties,

    Zahid Hassan Tushar, Adeleke Ademakinwa, Jianwu Wang, Zhibo Zhang, and Sanjay Purushotham, “Cloudunet: Adapt- ing unet for retrieving cloud properties,” in 2024 IEEE In- ternational Geoscience and Remote Sensing Symposium. 2024 IEEE International Geoscience and Remote Sensing ...

    2024

  8. [16]

    Transfer-learning-based approach to retrieve the cloud proper- ties using diverse remote sensing datasets,

    Jingwei Li, Feng Zhang, Wenwen Li, Xuan Tong, BaoXi- ang Pan, Jun Li, Han Lin, Husi Letu, and Frahan Mustafa, “Transfer-learning-based approach to retrieve the cloud proper- ties using diverse remote sensing datasets,”IEEE Transactions on Geoscience and Remote Sensing, 2023

    2023

  9. [17]

    Cloud identification and properties retrieval of the fengyun-4a satellite using a resunet model,

    Zhijun Zhao, Feng Zhang, Qiong Wu, Zhengqiang Li, Xuan Tong, Jingwei Li, and Wei Han, “Cloud identification and properties retrieval of the fengyun-4a satellite using a resunet model,” IEEE Transactions on Geoscience and Remote Sens- ing, vol. 61, pp. 1–18, 2023

    2023

  10. [18]

    Retrieval of cloud properties from thermal infrared radiometry using convolutional neural network,

    Quan Wang, Chen Zhou, Xiaoyong Zhuge, Chao Liu, Fuzhong Weng, and Minghuai Wang, “Retrieval of cloud properties from thermal infrared radiometry using convolutional neural network,” Remote Sensing of Environment , vol. 278, pp. 113079, 2022

    2022

  11. [19]

    Cloud identification and property retrieval from himawari-8 infrared measurements via a deep neural network,

    Xinyue Wang, Hironobu Iwabuchi, and Takaya Yamashita, “Cloud identification and property retrieval from himawari-8 infrared measurements via a deep neural network,” Remote Sensing of Environment, vol. 275, pp. 113026, 2022

    2022

  12. [20]

    Machine learning-based retrieval of day and night cloud macrophysical parameters over east asia using himawari- 8 data,

    Yikun Yang, Wenxiao Sun, Yulei Chi, Xing Yan, Hao Fan, Xingchuan Yang, Zhanshan Ma, Quan Wang, and Chuanfeng Zhao, “Machine learning-based retrieval of day and night cloud macrophysical parameters over east asia using himawari- 8 data,” Remote Sensing of Environment, vol. 273,...

    2022

  13. [21]

    The large-eddy simulation (les) atmospheric radiation measure- ment (arm) symbiotic simulation and observation (lasso) activ- ity for continental shallow convection,

    William I Gustafson, Andrew M V ogelmann, Zhijin Li, Xi- aoping Cheng, Kyle K Dumas, Satoshi Endo, Karen L John- son, Bhargavi Krishna, Tami Fairless, and Heng Xiao, “The large-eddy simulation (les) atmospheric radiation measure- ment (arm) symbiotic simulation and observation...

    2020

  14. [22]

    The spherical harmonics discrete ordinate method for three-dimensional atmospheric radiative transfer,

    K Franklin Evans, “The spherical harmonics discrete ordinate method for three-dimensional atmospheric radiative transfer,” Journal of the Atmospheric Sciences , vol. 55, no. 3, pp. 429– 446, 1998

    1998

  15. [23]

    Cbam: Convolutional block attention module,

    Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon, “Cbam: Convolutional block attention module,” in Proceedings of the European conference on computer vision (ECCV), 2018, pp. 3–19

    2018

  16. [24]

    Arbitrary style transfer in real-time with adaptive instance normalization,

    Xun Huang and Serge Belongie, “Arbitrary style transfer in real-time with adaptive instance normalization,” in Proceed- ings of the IEEE international conference on computer vision, 2017, pp. 1501–1510

    2017

  17. [25]

    Mul- tiple style transfer via variational autoencoder,

    Zhi-Song Liu, Vicky Kalogeiton, and Marie-Paule Cani, “Mul- tiple style transfer via variational autoencoder,” in 2021 IEEE International Conference on Image Processing (ICIP) . IEEE, 2021, pp. 2413–2417

    2021

  18. [26]

    3s-net: Arbitrary semantic- aware style transfer with controllable roi choice,

    Bingqing Guo and Pengwei Hao, “3s-net: Arbitrary semantic- aware style transfer with controllable roi choice,” in2021 IEEE International Conference on Image Processing (ICIP) . IEEE, 2021, pp. 2408–2412

    2021

  19. [27]

    Domain dilation for single domain generalization,

    Yuehui Fan, Baoyao Yang, Meng Shen, and Fei Lyu, “Domain dilation for single domain generalization,” in2024 IEEE Inter- national Conference on Image Processing (ICIP). IEEE, 2024, pp. 3931–3937

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.