Pith. sign in

REVIEW 3 major objections 5 minor 34 references

AsymLLIC: Asymmetric Lightweight Learned Image Compression

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read AsymLLIC shows that a learned image codec can keep its heavyweight encoder and still match VVC with a decoder that costs only 51.47 GMACs and 19.65M parameters.

desk verdict Solid engineering paper with a real asymmetric decoder contribution, but the missing TCM-small baseline makes its headline complexity claim underdetermined as written. read the letter →

arxiv 2412.17270 v1 pith:K2VGPJVJ submitted 2024-12-23 eess.IV

Shen Wang , Zhengxue Cheng , Donghui Feng , Guo Lu , Li Song , Wenjun Zhang This is my paper · ORCID
classification eess.IV
keywords learnedimagecompressionasymmetricencoder-decoderlightweightdecodertwo-stagetrainingrate-distortionoptimizationSwinTransformersimplificationhyperpriorcontextmodelcomplexity
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

Learned image codecs usually pair a heavy encoder with a heavy decoder, which makes decoding slow on low-end devices. This paper proposes AsymLLIC, a learned codec that deliberately shifts computational cost to the encoder and keeps the decoder light, using a two-stage training scheme that gradually swaps complex decoder modules for simpler ones. On Kodak, AsymLLIC reportedly beats BPG by 18.68% BD-rate while using only 51.47 GMACs and 19.65M parameters for decoding, and its rate-distortion curves sit close to VVC. If correct, this is the first learned codec to decouple encoder and decoder complexity to this degree, suggesting that efficient decoding on constrained devices does not have to sacrifice compression performance.

What carries the argument

The central mechanism is the asymmetric two-stage training strategy. In stage one, the synthesis decoder is swapped for a lighter structure and fine-tuned with a distortion-only loss while the encoder and entropy model are frozen; in stage two, the hyperprior decoder and context model are swapped for lighter versions and retrained together with the hyperprior encoder using the full rate-distortion loss. This staged substitution is what lets the asymmetric architecture hold onto RD performance while cutting decoder MACs from 145.71 in TCM to 51.47 in AsymLLIC.

What would settle it

Run the released model on a mobile CPU and measure per-image decode time against TCM under identical conditions; if AsymLLIC is not faster, or if the Kodak BD-rate cannot be reproduced near 18.68%, the central claim fails.

Watch

Extended reading notes

Core claim

The paper claims that the decoder side of a learned image codec can be drastically simplified without a proportionate loss in rate-distortion performance, provided the simplification is done with an asymmetric training schedule rather than by simply shrinking the network. In AsymLLIC, the complex synthesis decoder is first replaced by a lightweight version and fine-tuned alone with a distortion-only loss while the encoder and entropy model are frozen; then the hyperprior decoder and context model are replaced by lighter versions and retrained jointly with the hyperprior encoder under the full rate-distortion loss. The resulting model reports an 18.68% BD-rate improvement over BPG on Kodak with a decoder of 51.47 GMACs and 19.65M parameters, cutting decoder MACs to about one third of the TCM baseline while keeping RD performance close to VVC. The paper interprets this as evidence that computation can be effectively offloaded to the encoding side in learned image compression.

Load-bearing premise

The two-stage training procedure must converge to reconstruction quality as good as jointly training the full asymmetric network would; if the frozen encoder's latents are incompatible with the simplified decoder, the reported rate-distortion gains could degrade substantially.

Editorial extensions

If this is right

  • Decoder-side complexity in learned image codecs can be treated as a separate design variable, allowing a heavy encoder on a server and a light decoder on client devices.
  • The same two-stage training recipe can be applied to other symmetric learned codecs beyond TCM-small, potentially giving them similarly light decoders without retraining from scratch.
  • A decoder of 51.47 GMACs and 19.65M parameters brings learned image compression within reach of mobile and embedded hardware that could not run symmetric models like TCM.
  • The authors state the asymmetric design can be extended to learned video compression, which could make efficient video decoding on low-end devices feasible.

Reading between the lines

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

  • Because the stage-one synthesis decoder is fine-tuned with distortion only, that step may act as a form of knowledge distillation from the complex decoder; if so, the same gain could be obtained by direct distillation, which would be a cheaper alternative to the two-stage schedule.
  • The paper's MACs and parameter counts are theoretical; actual decoding latency on a given device also depends on memory bandwidth, kernel implementation, and quantization, so the practical speedup over TCM could differ from the 2.8x MAC ratio.
  • The reversed pyramid channel structure and removal of shifted window attention suggest that global receptive field matters less on the reconstruction path than on the analysis path; this asymmetry might transfer to other image restoration tasks.
  • The claim that the design methodology applies to 'any LIC model' is broader than what the experiments show, since only TCM-small is tested; verifying it on stronger baselines like full TCM or transformer-only codecs would test the generalization.
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

3 major / 5 minor

Summary. The paper proposes AsymLLIC, a learned image compression architecture with an asymmetric encoder-decoder design that moves computational load to the encoder and uses a lightweight decoder. The method is built on TCM-small and introduces a two-stage training strategy in which the synthesis decoder is first fine-tuned with a distortion-only loss while the encoder and entropy model are frozen, and then the hyperprior decoder and context model are lightweighted and trained together with the hyperprior encoder. The authors report a BD-rate of -18.68% over BPG on Kodak with 19.65M decoder parameters and 51.47 GMACs, claim near-VVC performance, and include ablations on the synthesis decoder and hyperprior path structures.

Significance. If the reported complexity-performance trade-off is accurate, AsymLLIC is a useful contribution to practical learned image compression: it demonstrates that decoding complexity can be substantially reduced while retaining RD performance near traditional codecs, and the two-stage training idea is simple and potentially transferable to other LIC models. The paper evaluates on three datasets and includes ablation studies that isolate design choices. However, the verification is incomplete: the direct baseline TCM-small is omitted from the main complexity comparison, the staged training procedure is not ablated against joint training, and the small performance differences in the ablations are not accompanied by variance estimates or released code/weights. These gaps currently prevent the reader from verifying the central claim that AsymLLIC achieves the 'optimal balance' between compression performance and decoding complexity.

major comments (3)
  1. [Section III-B, Table I] Table I omits TCM-small, which Section III-B identifies as the base architecture that AsymLLIC starts from. Since TCM-small's synthesis-decoder complexity is already plotted in Figure 3(e), the authors demonstrably have the data to include a TCM-small row. Without this row, the central claim of 'lowest decoding complexity' and 'optimal balance' cannot be verified: if TCM-small has comparable or lower decoder MACs, or better BD-rate at similar complexity, the claimed advantage over previous LIC methods is overstated. Please add TCM-small (with the same evaluation protocol and input size) to Table I and discuss the comparison explicitly.
  2. [Section III-A, Figure 2] The two-stage asymmetric training strategy is not validated. Stage 1 freezes the encoder and entropy model and trains only the lightweight synthesis decoder with a distortion-only loss; stage 2 then trains the hyperprior decoder and context model together with ha. The paper provides no ablation comparing this staged procedure with (i) joint end-to-end training of the full asymmetric network, or (ii) fine-tuning all decoder modules simultaneously from the TCM-small checkpoint. Without such an ablation, the reader cannot tell whether the staged training is necessary or whether it converges to a worse optimum than joint training. Please add an ablation that isolates the effect of the training scheme on the final RD performance.
  3. [Tables II and III] The reported performance differences are small (e.g., 0.18 dB in Table II) and no variance or statistical significance is reported. Since the paper does not release code or trained weights, the reader cannot assess whether the proposed structure's advantage over the RBU baseline is within run-to-run variability. Please report the standard deviation across at least three independent training runs, or release trained models, to support the claimed superiority of the proposed decoder structures.
minor comments (5)
  1. [Abstract] The word 'evitably' should be 'inevitably'.
  2. [Figure 2] The caption of Figure 2(d) contains the typo 'Aymmetric'; it should be 'Asymmetric'.
  3. [Table I caption] The caption has a typo: 'comparsion' should be 'comparison'.
  4. [Figure 4] The legend lists Balle18, Minnen18, Cheng20, Xie21, Qian22, TinyLIC, and Ours, but the text claims comparisons with TCM; clarify whether TCM is included in Figure 4 and, if so, why it is missing from the legend.
  5. [Section IV-C] The ablation for the hyperprior path (Table III) varies the number of slices and the structure simultaneously; it would be clearer to state which factors are held fixed when comparing 'slice number = 5, Ours' with the other rows.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the claimed BD-rate and decoder-complexity numbers are external benchmark measurements, not derived from the method's assumptions.

full rationale

The paper is an empirical architecture study. The central claim (18.68% BD-rate improvement over BPG on Kodak with 51.47 GMACs and 19.65M decoder parameters) is a measured outcome of training and evaluation against external codecs (BPG, VVC) and prior LIC models, not a quantity obtained by construction from an input parameter. The asymmetric training strategy in Section III-A is a staged optimization procedure, and its effectiveness is assessed by final rate-distortion curves and ablations (Tables II and III); no equation in the paper defines the final BD-rate as a function of the training assumptions, so there is no self-definitional reduction. The base architecture is TCM from Liu et al. [11], which is external to this paper, and the self-citations to Cheng20 [7] and [21] appear only as related-work baselines and are not load-bearing for the new contribution. The absence of an explicit TCM-small row in Table I is a reporting or comparison completeness gap, but it does not make the reported numbers circular because the BD-rate and MACs are independently measured rather than derived from the omitted baseline. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no ansatz is smuggled in via citation. The staged-training convergence concern raised by the reader is an empirical risk, not a circularity: the paper tests and reports the final RD performance rather than assuming it. Overall, the derivation chain is self-contained with respect to external benchmarks, so the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claims rest on standard LIC training assumptions plus one paper-specific premise (two-stage warm-up training), and on hand-chosen architecture hyperparameters (slice count, channel widths). No new physical or conceptual entities are introduced.

free parameters (3)
  • number of slices in channel-wise context model = 5
    Selected by ablation in Table III; increasing slices from 1 to 5 lowers bpp from 0.412 to 0.324 at similar MACs.
  • reversed pyramid channel widths = not fully specified
    Chosen by hand to keep decoder MACs near 35 GMACs for the synthesis decoder; exact per-layer widths are not reported.
  • RD tradeoff lambdas = {0.0025, 0.0035, 0.0067, 0.0130, 0.0250, 0.0500}
    Standard rate-distortion operating point sweep; not an ad hoc parameter for the core claim.
assumptions (4)
  • domain assumption The rate-distortion loss L = R + lambda D is the correct objective for optimizing compression quality.
    Section IV-A uses L = R + lambda D throughout; this is the standard LIC training objective but is an assumption about what quality means.
  • domain assumption TCM-small is a valid strong base architecture and the proposed simplifications preserve its encoder-side representation.
    The method builds on TCM-small (Section III-B) without re-justifying the base design.
  • ad hoc to paper The two-stage training scheme converges to a good local optimum of the final RD objective.
    Section III-A: stage 1 trains g_s^s with distortion-only loss while freezing other modules; the paper provides no proof or analysis that this staged procedure reaches the same quality as joint training.
  • domain assumption MACs measured at 768x512 input are a fair proxy for decoder complexity across compared methods.
    Table I reports Dec MACs using 768x512 input; different methods may have different input scaling behaviors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AsymLLIC: Asymmetric Lightweight Learned Image Compression." pith.science (2026). https://pith.science/paper/K2VGPJVJ

@misc{pith2026241217270,
  author       = {Pith},
  title        = {Pith review of: AsymLLIC: Asymmetric Lightweight Learned Image Compression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/K2VGPJVJ}},
  note         = {Machine review of arXiv:2412.17270}
}
read the original abstract

Learned image compression (LIC) methods often employ symmetrical encoder and decoder architectures, evitably increasing decoding time. However, practical scenarios demand an asymmetric design, where the decoder requires low complexity to cater to diverse low-end devices, while the encoder can accommodate higher complexity to improve coding performance. In this paper, we propose an asymmetric lightweight learned image compression (AsymLLIC) architecture with a novel training scheme, enabling the gradual substitution of complex decoding modules with simpler ones. Building upon this approach, we conduct a comprehensive comparison of different decoder network structures to strike a better trade-off between complexity and compression performance. Experiment results validate the efficiency of our proposed method, which not only achieves comparable performance to VVC but also offers a lightweight decoder with only 51.47 GMACs computation and 19.65M parameters. Furthermore, this design methodology can be easily applied to any LIC models, enabling the practical deployment of LIC techniques.

Figures

Figures reproduced from arXiv: 2412.17270 by the authors.

Figure 1
Figure 1. Performance versus decoder complexity on Kodak dataset. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overall framework of our asymmetric structure and asymmetric training strategy. The snowflake pattern indicates that the module [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The designed architecture of synthesis decoder. Subfig (a) and (b) show the original structures of TCM-small and our proposed [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: R-D performance of traditional codec and LIC methods evaluated on different datasets. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 25 canonical work pages

  1. [1]

    The jpeg still picture compression standard,

    Gregory K Wallace, “The jpeg still picture compression standard,” Communications of the ACM , vol. 34, no. 4, pp. 30–44, 1991

    work page 1991

  2. [2]

    An overview of the jpeg 2000 still image compression standard,

    Majid Rabbani and Rajan Joshi, “An overview of the jpeg 2000 still image compression standard,” Signal processing: Image communication, vol. 17, no. 1, pp. 3–48, 2002

    2000

  3. [3]

    Overview of the high efficiency video coding (hevc) standard,

    Gary J Sullivan, Jens-Rainer Ohm, Woo-Jin Han, and Thomas Wiegand, “Overview of the high efficiency video coding (hevc) standard,” IEEE Transactions on circuits and systems for video technology , vol. 22, no. 12, pp. 1649–1668, 2012

    work page 2012

  4. [4]

    Versatile video coding– towards the next generation of video compression,

    Jens-Rainer Ohm and Gary J Sullivan, “Versatile video coding– towards the next generation of video compression,” in Picture Coding Symposium, 2018, vol. 2018

    work page 2018

  5. [5]

    Variational image compression with a scale hyperprior,

    Johannes Ball ´e, David Minnen, Saurabh Singh, Sung Jin Hwang, and Nick Johnston, “Variational image compression with a scale hyperprior,” arXiv preprint arXiv:1802.01436 , 2018

    arXiv 2018

  6. [6]

    Joint au- toregressive and hierarchical priors for learned image compression,

    David Minnen, Johannes Ball ´e, and George D Toderici, “Joint au- toregressive and hierarchical priors for learned image compression,” Advances in neural information processing systems , vol. 31, 2018

    2018

  7. [7]

    Learned image compression with discretized gaussian mixture like- lihoods and attention modules,

    Zhengxue Cheng, Heming Sun, Masaru Takeuchi, and Jiro Katto, “Learned image compression with discretized gaussian mixture like- lihoods and attention modules,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 7939– 7948

    work page 2020

  8. [8]

    Enhanced invertible encoding for learned image compression,

    Yueqi Xie, Ka Leong Cheng, and Qifeng Chen, “Enhanced invertible encoding for learned image compression,” in Proceedings of the 29th ACM international conference on multimedia , 2021, pp. 162–170

    work page 2021

Show all 34 references
  1. [9]

    Entro- former: A transformer-based entropy model for learned image compres- sion,

    Yichen Qian, Ming Lin, Xiuyu Sun, Zhiyu Tan, and Rong Jin, “Entro- former: A transformer-based entropy model for learned image compres- sion,” arXiv preprint arXiv:2202.05492 , 2022

    2022 arXiv

  2. [10]

    High-efficiency lossy image coding through adaptive neighborhood information aggre- gation,

    Ming Lu, Fangdong Chen, Shiliang Pu, and Zhan Ma, “High-efficiency lossy image coding through adaptive neighborhood information aggre- gation,” arXiv preprint arXiv:2204.11448 , 2022

    2022 arXiv

  3. [11]

    Learned image compression with mixed transformer-cnn architectures,

    Jinming Liu, Heming Sun, and Jiro Katto, “Learned image compression with mixed transformer-cnn architectures,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 14388–14397

    2023

  4. [12]

    End-to-end optimized image compression,

    Johannes Ball ´e, Valero Laparra, and Eero P Simoncelli, “End-to-end optimized image compression,” arXiv preprint arXiv:1611.01704 , 2016

    2016 arXiv

  5. [13]

    Channel-wise autoregressive entropy models for learned image compression,

    David Minnen and Saurabh Singh, “Channel-wise autoregressive entropy models for learned image compression,” in 2020 IEEE International Conference on Image Processing (ICIP) . IEEE, 2020, pp. 3339–3343

    2020

  6. [14]

    Nonlinear transforms in learned image compression from a communication perspective,

    Youneng Bao, Fanyang Meng, Chao Li, Siwei Ma, Yonghong Tian, and Yongsheng Liang, “Nonlinear transforms in learned image compression from a communication perspective,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 33, no. 4, pp. 1922–1936, 2022

    1922

  7. [15]

    Taylor series based dual-branch transformation for learned image compression,

    Youneng Bao, Wen Tan, Linfeng Zheng, Fanyang Meng, Wei Liu, and Yongsheng Liang, “Taylor series based dual-branch transformation for learned image compression,” Signal Processing , vol. 212, pp. 109128, 2023

    2023

  8. [16]

    Mobilenets: Efficient convolutional neural networks for mobile vision applications,

    Andrew G Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, and Hartwig Adam, “Mobilenets: Efficient convolutional neural networks for mobile vision applications,” arXiv preprint arXiv:1704.04861 , 2017

    2017 arXiv

  9. [17]

    Fast and high-performance learned image compres- sion with improved checkerboard context model, deformable residual module, and knowledge distillation,

    Haisheng Fu, Feng Liang, Jie Liang, Yongqiang Wang, Guohe Zhang, and Jingning Han, “Fast and high-performance learned image compres- sion with improved checkerboard context model, deformable residual module, and knowledge distillation,” arXiv preprint arXiv:2309.02529 , 2023

    2023 arXiv

  10. [18]

    Context- adaptive entropy model for end-to-end optimized image compression,

    Jooyoung Lee, Seunghyun Cho, and Seung-Kwon Beack, “Context- adaptive entropy model for end-to-end optimized image compression,” arXiv preprint arXiv:1809.10452 , 2018

    2018 arXiv

  11. [19]

    Joint global and local hierarchical priors for learned image compression,

    Jun-Hyuk Kim, Byeongho Heo, and Jong-Seok Lee, “Joint global and local hierarchical priors for learned image compression,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 5992–6001

    2022

  12. [20]

    Deep image compression with residual learning,

    Weigui Li, Wenyu Sun, Yadong Zhao, Zhuqing Yuan, and Yongpan Liu, “Deep image compression with residual learning,” Applied Sciences , vol. 10, no. 11, pp. 4023, 2020

    2020

  13. [21]

    Deep residual learning for image compression.,

    Zhengxue Cheng, Heming Sun, Masaru Takeuchi, and Jiro Katto, “Deep residual learning for image compression.,” in Cvpr workshops , 2019, p. 0

    2019

  14. [22]

    Neural image compression via attentional multi- scale back projection and frequency decomposition,

    Ge Gao, Pei You, Rong Pan, Shunyuan Han, Yuanyuan Zhang, Yuchao Dai, and Hojae Lee, “Neural image compression via attentional multi- scale back projection and frequency decomposition,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 14677–14686

    2021

  15. [23]

    Wavelet- based deep auto encoder-decoder (wdaed)-based image compression,

    Dipti Mishra, Satish Kumar Singh, and Rajat Kumar Singh, “Wavelet- based deep auto encoder-decoder (wdaed)-based image compression,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 31, no. 4, pp. 1452–1462, 2020

    2020

  16. [24]

    Expanded adaptive scaling normalization for end to end image compression,

    Chajin Shin, Hyeongmin Lee, Hanbin Son, Sangjin Lee, Dogyoon Lee, and Sangyoun Lee, “Expanded adaptive scaling normalization for end to end image compression,” in European Conference on Computer Vision . Springer, 2022, pp. 390–405

    2022

  17. [25]

    Joint graph attention and asymmetric convolutional neural network for deep image compression,

    Zhisen Tang, Hanli Wang, Xiaokai Yi, Yun Zhang, Sam Kwong, and C- C Jay Kuo, “Joint graph attention and asymmetric convolutional neural network for deep image compression,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 33, no. 1, pp. 421–433, 2022

    2022

  18. [26]

    The devil is in the details: Window-based attention for image compression,

    Renjie Zou, Chunfeng Song, and Zhaoxiang Zhang, “The devil is in the details: Window-based attention for image compression,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 17492–17501

    2022

  19. [27]

    Shufflenet: An extremely efficient convolutional neural network for mobile devices,

    Xiangyu Zhang, Xinyu Zhou, Mengxiao Lin, and Jian Sun, “Shufflenet: An extremely efficient convolutional neural network for mobile devices,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 6848–6856

    2018

  20. [28]

    Ghostnet: More features from cheap operations,

    Kai Han, Yunhe Wang, Qi Tian, Jianyuan Guo, Chunjing Xu, and Chang Xu, “Ghostnet: More features from cheap operations,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 1580–1589

    2020

  21. [29]

    Ghostnets on heterogeneous devices via cheap operations,

    Kai Han, Yunhe Wang, Chang Xu, Jianyuan Guo, Chunjing Xu, Enhua Wu, and Qi Tian, “Ghostnets on heterogeneous devices via cheap operations,” International Journal of Computer Vision , vol. 130, no. 4, pp. 1050–1069, 2022

    2022

  22. [30]

    Efficient lightweight attention based learned image compression,

    Ziyang He, Lei Luo, Le Zhang, Hongwei Guo, and Ce Zhu, “Efficient lightweight attention based learned image compression,” in 2023 IEEE International Conference on Visual Communications and Image Processing (VCIP). IEEE, 2023, pp. 1–5

    2023

  23. [31]

    Theo- retical bound-guided hierarchical vae for neural image codecs,

    Yichi Zhang, Zhihao Duan, Yuning Huang, and Fengqing Zhu, “Theo- retical bound-guided hierarchical vae for neural image codecs,” arXiv preprint arXiv:2403.18535, 2024

    2024 arXiv

  24. [32]

    Leveraging redundancy in feature for efficient learned image compression,

    Peng Qin, Youneng Bao, Fanyang Meng, Wen Tan, Chao Li, Genhong Wang, and Yongsheng Liang, “Leveraging redundancy in feature for efficient learned image compression,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 20...

    2024

  25. [33]

    A large-scale hierarchical image database,

    Jia Deng, “A large-scale hierarchical image database,” Proc. of IEEE Computer Vision and Pattern Recognition, 2009 , 2009

    2009

  26. [34]

    Adam: A method for stochastic optimization,

    Diederik P Kingma and Jimmy Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.