Pith. sign in

REVIEW 4 major objections 5 minor 46 references

DT-JRD: Deep Transformer based Just Recognizable Difference Prediction Model for Video Coding for Machines

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read A single forward pass on the original image can predict the distortion level at which machine vision starts to fail, and using that prediction to guide the VVC codec saves about 29.58% of the bit rate while preserving object-detection…

desk verdict A credible single-image JRD predictor for VCM, with a coding-gain claim that is plausible but not yet pinned down because the background-QP choice is unstated and untested. read the letter →

arxiv 2411.09308 v1 pith:Y3OMVJUG submitted 2024-11-14 eess.IV cs.CV

classification eess.IVcs.CV
keywords JustRecognizableDifferenceVideoCodingforMachinesobjectdetectionVisionTransformerGaussiansoftlabelsVVCbit-ratereductionmulti-classclassification
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 tries to establish that the Just Recognizable Difference (JRD)—the smallest amount of compression distortion that makes a machine-vision system's output change—can be predicted directly from a single original image, and that the prediction is accurate enough to drive a Video Coding for Machines (VCM) system. The proposed DT-JRD model casts JRD prediction as a 64-class classification task solved by a fine-tuned Vision Transformer, instead of the previous approach of running many binary classifiers on many pre-compressed versions of the same image. On the Object-Wise JRD dataset the model reports a mean absolute error of 5.574, about 13.1% better than the best previous predictor. When the predicted JRD is used to set per-block quantization in the Versatile Video Coding (VVC) standard, the authors report an average 29.58% bit-rate reduction at matched object-detection accuracy. If true, this turns a costly multi-pass search into a one-pass estimate and gives VCM a practical way to spend bits only where the machine vision task needs them.

What carries the argument

The load-bearing mechanism is the DT-JRD network itself: a Vision Transformer (ViT-Large/32) that takes only the original object image and outputs a probability distribution over 64 VVC distortion levels. Three design choices carry the result: (1) the Distortion-Aware Fine-Tuning (DAFT) strategy, which freezes the patch embedding, class token, position embedding, and final layer norm while training the transformer blocks and classification head, so the pretrained model adapts without overfitting on the small JRD dataset; (2) patch-token average pooling rather than the class token for the classification head, because pooled patch features carry more JRD-relevant distortion information; and (3) the Gaussian Distribution-based Soft Label (GDSL) loss, which replaces one-hot labels with a Gaussian centered at the true JRD and standard deviation σ=3, effectively augmenting labels with neighboring distortion levels and relaxing the classification boundary.

What would settle it

Run the DT-JRD VCM pipeline on a benchmark where background context matters for detection—for instance, small or partially occluded objects whose surroundings provide essential cues—and compare the mAP@0.5 versus bit-rate curve against VVC all-intra; if the curve dips below VVC at matched bit rates, the background-QP assumption fails.

Watch

Extended reading notes

Core claim

The core discovery is that JRD is a learnable, content-dependent threshold, and a transformer trained with soft Gaussian labels can predict it in one forward pass. The paper shows that neighboring distortion levels around the ground-truth JRD are visually and semantically close, so instead of treating only the exact JRD class as correct, the loss assigns partial probability to nearby classes according to a Gaussian centered at the true JRD with standard deviation σ = 3. Combined with a Distortion-Aware Fine-Tuning strategy and 2D-interpolated position embeddings for 384×384 inputs, this brings the predicted-JRD error down to 5.574 MAE. Feeding the predicted JRD into a VVC all-intra encoder as a per-object quantization constraint—background coding tree units get a larger QP than object regions—yields an average 29.58% bit-rate reduction at matched mAP@0.5 and 26.84% at mAP@0.75 compared with unmodified VVC, while the detector's accuracy is maintained.

Load-bearing premise

The load-bearing assumption is that a detector's accuracy on an object depends only on the content inside its bounding box, so the surrounding background can be compressed much more aggressively; the paper tests this with one detector and one data split rather than establishing it as a general property of machine vision.

Editorial extensions

If this is right

  • A single source image is enough to predict JRD, so the expensive N-times compression-and-classification loop of earlier predictors can be dropped.
  • Because the predicted JRD is about 1.67 QP steps above the ground truth on average, the measured bit-rate savings are conservative and could grow with a better-calibrated predictor.
  • Per-CTU quantization guided by predicted JRD transfers the gain to any standard encoder: the paper demonstrates a 29.58% bit-rate saving at matched mAP@0.5 over VVC all-intra.
  • The same predicted-JRD interface can be plugged into other coding pipelines without changing the decoder or the downstream object detector.

Reading between the lines

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

  • An untested extension is to feed the predicted JRD into a learned image or video codec as an adaptive quality constraint rather than mapping it to VVC QP; the paper only demonstrates the VVC path.
  • The background-QP assumption is validated on one detector and one data split; on detectors that use context outside the bounding box, the reported 29.58% saving may shrink because the detector would notice the coarser background.
  • The GDSL idea is not JRD-specific: any ordered-label prediction task where nearby classes are semantically close, such as bitrate ladders or difficulty ratings, could benefit from the same Gaussian soft-label loss.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 DT-JRD, a ViT-based multi-class classifier that predicts object-wise Just Recognizable Difference (JRD) directly from the original image, avoiding the multi-pass binary-classification search of prior JRD predictors. Training uses Gaussian Distribution-based Soft Labels (GDSL), and the predicted JRD is used to guide VVC all-intra coding by encoding object CTUs at the predicted JRD QP and background CTUs at a larger QP. The authors report a JRD prediction MAE of 5.574, a 13.1% improvement over BC-JRD, and a 29.58% BDBR reduction against VVC all-intra at matched mAP for Faster R-CNN object detection.

Significance. If confirmed, the single-pass prediction is a practical advantage over prior multi-pass JRD predictors, and the GDSL soft-label formulation is a simple, transferable idea for ordinal classification near perceptual thresholds. The paper includes comparisons to two baselines, ablations of learning strategy, output features, and position embedding, and a wide set of accuracy metrics (MAE, PSNR, SSIM, bit rate, R2). The main uncertainty is not the JRD prediction itself but the coding-gain evaluation, which currently rests on an unspecified background-QP choice and a nonstandard mAP protocol; these issues are fixable with additional experiments and therefore warrant major revision rather than rejection.

major comments (4)
  1. [Section III-A.3] The mAP is computed using the detections of Faster R-CNN on the original images as ground truth rather than the dataset's human annotations. This protocol measures only consistency with the uncompressed detector output and can hide detection failures on objects that the detector already misses in the original image. Since the central coding claim is "maintaining object detection accuracy," please recompute mAP against COCO ground-truth boxes, or justify the pseudo-ground-truth choice and show that the BDBR conclusions are unchanged.
  2. [Section II-E and Section III-D] The background CTU QP, denoted QPb, is a free parameter of the proposed VCM, but the paper never states its value, how it was chosen, or how sensitive the 29.58% BDBR saving is to it. Because background CTUs dominate the bit budget, the reported coding gain could be driven by QPb rather than by the predicted JRD. Please specify the QPb values used, add an ablation over QPb (including a foreground-only baseline without any background QP increase), and report a foreground/background bit-rate decomposition.
  3. [Section III-D and Section III-A.1] The coding experiments use a single detector (Faster R-CNN), a single random 8:1:1 split, and no confidence intervals or repeated runs. Since the JRD ground truth was collected on cropped object images, transferring object-wise JRD thresholds to full-image CTU-level coding is an extrapolation. Please add at least one additional detector (e.g., YOLO or DETR), evaluate on multiple splits, and report mean and standard deviation for MAE and BDBR; otherwise the generality of the 29.58% claim is not established.
  4. [Section II-D, Table I, and Section III-B] The reported MAE values are drawn from different evaluation sets (validation in Table I, category-specific and overall test values in Table II, and ablation values in Table III), but the table captions and surrounding text do not consistently say which set each number refers to. Please state the dataset split explicitly for every table and metric, and clarify that the abstract's 5.574 is the Table III ablation value while the full-test overall EA in Table II is 5.57 after rounding.
minor comments (5)
  1. [Table II] The column order in Table II is confusing because the header repeats "EA" and the text refers to "the first five columns" and "the last two columns." Please label the columns explicitly as Small, Medium, Large, Person, Car, EA, and E[27,51].
  2. [Section III-A.3 and Section III-D] The QP settings are inconsistent: Section III-A.3 states that VVC all-intra uses QP ∈ {25, 27, 29, 31, 33}, while Section III-D states that the QP of the original VVC all-intra is from 28 to 35. Please unify the settings and report the exact operating points used for each BDBR calculation.
  3. [Section III-D] The sentence "The predicted JRD by the DT-JRD is averagely larger than the ground truth JRD for about 1.67" should be tied to the QP-offset analysis; please clarify whether this bias contributes to the bit-rate savings and whether it affects the mAP comparison.
  4. [Table III] The checkmark-only notation in Table III makes the configurations hard to read; please use explicit Yes/No entries or separate rows for each configuration so that the reader can identify exactly which modules are active in each ablation.
  5. [General] No code, pretrained models, or evaluation scripts are released. Given the need for replication of the VCM results, a public release would strengthen the paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DT-JRD is a supervised predictor evaluated on held-out hard labels, and the VCM coding gain is benchmarked against VVC all-intra.

full rationale

DT-JRD is trained and evaluated as a supervised multi-class classifier on the OW-JRD dataset [37]. The ground-truth JRD labels are external measurements (object-wise detection thresholds from VVC-coded distortions), not outputs of the proposed model; evaluation uses a held-out test split (Section III-A.1) with MAE computed against these labels. The GDSL soft labels are derived from the same ground-truth labels but only as a training-time label-smoothing device; the reported test MAE is computed against hard labels, so the loss design does not define the evaluation metric. The VCM bit-rate claim is measured against VVC all-intra and against GT-JRD, BC-JRD, and EL-JRD coding curves (Figs. 9 and 10); BDBR and BD-mAP are external rate-accuracy comparisons, not fitted quantities. The only validation-set choice is the Gaussian width sigma in Table I, which is a hyperparameter, not a definitional identity. The paper does not specify background QPb and tests only one detector, but that is a missing generality check, not circularity. Self-citations to [37] provide the dataset and the BC-JRD baseline, yet the present evaluation does not reduce to those citations because the reported numbers are benchmark measurements against held-out labels and an external codec.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper's central claims rest on three domain assumptions: label quality, smooth machine response near JRD, and the background-CTU compression assumption. The only fitted numeric parameter is the GDSL width sigma. No new physical entities are introduced.

free parameters (3)
  • GDSL standard deviation sigma = 3
    Gaussian width in Eq. 4, selected on validation set from {2,...,7} in Table I; changes loss shape and final predictions.
  • Input image resolution = 384x384
    Fine-tuning resolution; authors choose higher resolution than pretraining (224x224) to improve accuracy, ablated in Table III rows 3 vs 5.
  • DAFT layer freezing scheme = patch embedding, class token, position embedding, final LayerNorm frozen; feature extraction module and classification…
    Architectural choice justified by ablation (Table III rows 1-3); affects generalization on the small JRD dataset.
assumptions (5)
  • domain assumption The OW-JRD dataset labels are reliable ground-truth machine-vision thresholds.
    The whole supervised training and evaluation treat labels from reference 37 as correct JRD values; noise or detector-specific bias in labels would propagate to MAE and coding claims.
  • domain assumption Machine vision responses to distortion are smooth near the JRD, justifying Gaussian soft labels.
    Motivates GDSL loss in Section II-D, supported only by SSIM/LPIPS statistics on three sample images; if false, soft labels would blur away genuine threshold information.
  • domain assumption Background CTUs can be heavily compressed without reducing object-detection accuracy.
    Section II-E encodes background with QPb larger than predicted JRD; Faster R-CNN can use contextual cues, so this is load-bearing for the 29.58% bitrate saving claim.
  • standard math Standard cross-entropy and Gaussian normalization (Eq. 4).
    Used without proof; uncontroversial.
  • ad hoc to paper Fine-tuning only deep layers of a pretrained ViT preserves useful representations and avoids overfitting.
    DAFT strategy in Section II-C, justified by ablation rather than a theorem.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DT-JRD: Deep Transformer based Just Recognizable Difference Prediction Model for Video Coding for Machines." pith.science (2026). https://pith.science/paper/Y3OMVJUG

@misc{pith2026241109308,
  author       = {Pith},
  title        = {Pith review of: DT-JRD: Deep Transformer based Just Recognizable Difference Prediction Model for Video Coding for Machines},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y3OMVJUG}},
  note         = {Machine review of arXiv:2411.09308}
}
read the original abstract

Just Recognizable Difference (JRD) represents the minimum visual difference that is detectable by machine vision, which can be exploited to promote machine vision oriented visual signal processing. In this paper, we propose a Deep Transformer based JRD (DT-JRD) prediction model for Video Coding for Machines (VCM), where the accurately predicted JRD can be used reduce the coding bit rate while maintaining the accuracy of machine tasks. Firstly, we model the JRD prediction as a multi-class classification and propose a DT-JRD prediction model that integrates an improved embedding, a content and distortion feature extraction, a multi-class classification and a novel learning strategy. Secondly, inspired by the perception property that machine vision exhibits a similar response to distortions near JRD, we propose an asymptotic JRD loss by using Gaussian Distribution-based Soft Labels (GDSL), which significantly extends the number of training labels and relaxes classification boundaries. Finally, we propose a DT-JRD based VCM to reduce the coding bits while maintaining the accuracy of object detection. Extensive experimental results demonstrate that the mean absolute error of the predicted JRD by the DT-JRD is 5.574, outperforming the state-of-the-art JRD prediction model by 13.1%. Coding experiments shows that comparing with the VVC, the DT-JRD based VCM achieves an average of 29.58% bit rate reduction while maintaining the object detection accuracy.

Figures

Figures reproduced from arXiv: 2411.09308 by the authors.

Figure 2
Figure 2. 2D Interpolation for Position Embedding. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The framework of the proposed DT-JRD prediction model. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 5
Figure 5. Relationship between the distance to the JRD and the similarity to the JRD image [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figures from the paper (6 more)
Figure 6
Figure 6. Figure 6: Visualization for different types of labels and their respective advantages and limitations. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: The processing workflow for DT-JRD based VCM optimization. [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Correlation between predicted JRD and ground truth JRD in terms of PSNR. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 10
Figure 10. Figure 10: Relationship between mAP@0.75 and average bit rate for different coding [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 11
Figure 11. Figure 11: The relationship between mAP@0.5 and bit rate for the proposed DT-JRD based [PITH_FULL_IMAGE:figures/full_fig_p010_11.png]
Figure 12
Figure 12. Figure 12: Visualized object detection results of compressed images from different coding methods. (a)(g)(m) original images, (b)(h)(n) VVC all-intra, (c)(i)(o) GT-JRD based VCM, [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 36 canonical work pages

  1. [1]

    Faster r-cnn: Towards real-time object detection with region proposal networks,

    S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal networks,” IEEE Trans. Pattern Anal. Mach. Intell. , vol. 39, no. 6, pp. 1137–1149, 2017

  2. [2]

    Video coding for machines: A paradigm of collaborative compression and intelligent analytics,

    L. Duan, J. Liu, W. Yang, T. Huang, and W. Gao, “Video coding for machines: A paradigm of collaborative compression and intelligent analytics,” IEEE Trans. Image Process. , vol. 29, pp. 8680–8695, 2020

  3. [3]

    The jpeg still picture compression standard,

    G. Wallace, “The jpeg still picture compression standard,” IEEE Trans. Consum. Electron., vol. 38, no. 1, pp. xviii–xxxiv, 1992

  4. [4]

    The jpeg 2000 still image compression standard,

    A. Skodras, C. Christopoulos, and T. Ebrahimi, “The jpeg 2000 still image compression standard,” IEEE Signal Process. Mag., vol. 18, no. 5, pp. 36–58, 2001

  5. [5]

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

    G. J. Sullivan, J.-R. Ohm, W.-J. Han, and T. Wiegand, “Overview of the high efficiency video coding (hevc) standard,” IEEE Trans. Circuit Syst. Video Technol., vol. 22, no. 12, pp. 1649–1668, 2012

  6. [6]

    Overview of the versatile video coding (vvc) standard and its applications,

    B. Bross, Y .-K. Wang, Y . Ye, S. Liu, J. Chen, G. J. Sullivan, and J.-R. Ohm, “Overview of the versatile video coding (vvc) standard and its applications,” IEEE Trans. Circuit Syst. Video Technol., vol. 31, no. 10, pp. 3736–3764, 2021

  7. [7]

    End-to- end learnt image compression via non-local attention optimization and improved context modeling,

    T. Chen, H. Liu, Z. Ma, Q. Shen, X. Cao, and Y . Wang, “End-to- end learnt image compression via non-local attention optimization and improved context modeling,” IEEE Trans. Image Process. , vol. 30, pp. 3179–3191, 2021

  8. [8]

    Elic: Efficient learned image compression with unevenly grouped space- channel contextual adaptive coding,

    D. He, Z. Yang, W. Peng, R. Ma, H. Qin, and Y . Wang, “Elic: Efficient learned image compression with unevenly grouped space- channel contextual adaptive coding,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 2022, pp. 5708–5717

Show all 46 references
  1. [9]

    Mlic: Multi- reference entropy model for learned image compression,

    W. Jiang, J. Yang, Y . Zhai, P. Ning, F. Gao, and R. Wang, “Mlic: Multi- reference entropy model for learned image compression,” in Proc. ACM Int. Conf. Multimedia , 2023, pp. 7618–7627

  2. [10]

    Toward top-down just noticeable difference estimation of natural images,

    Q. Jiang, Z. Liu, S. Wang, F. Shao, and W. Lin, “Toward top-down just noticeable difference estimation of natural images,” IEEE Trans. Image Process., vol. 31, pp. 3697–3712, 2022

  3. [11]

    Perceptual image compression with block-level just noticeable difference prediction,

    T. Tian, H. Wang, S. Kwong, and C.-C. J. Kuo, “Perceptual image compression with block-level just noticeable difference prediction,” ACM Trans. Multimedia Comput. Commun. Appl. , vol. 16, no. 4, pp. 1–15, 2021

  4. [12]

    Bl- juniper: A cnn-assisted framework for perceptual video coding leverag- ing block-level jnd,

    S. Nami, F. Pakdaman, M. R. Hashemi, and S. Shirmohammadi, “Bl- juniper: A cnn-assisted framework for perceptual video coding leverag- ing block-level jnd,” IEEE Trans. Multimedia , vol. 25, pp. 5077–5092, 2023. IEEE TRANSACTIONS ON MULTIMEDIA 11 (a) (b) (c) (d) (e) (f) (g) (h...

  5. [13]

    Just noticeable difference level prediction for perceptual image compression,

    T. Tian, H. Wang, L. Zuo, C.-C. J. Kuo, and S. Kwong, “Just noticeable difference level prediction for perceptual image compression,” IEEE Trans. Broadcast., vol. 66, no. 3, pp. 690–700, 2020

  6. [14]

    Just noticeable distortion profile inference: A patch-level structural visibility learning approach,

    X. Shen, Z. Ni, W. Yang, X. Zhang, S. Wang, and S. Kwong, “Just noticeable distortion profile inference: A patch-level structural visibility learning approach,” IEEE Trans. Image Process. , vol. 30, pp. 26–38, 2021

  7. [15]

    Deep learning-based picture-wise just noticeable distortion prediction model for image compression,

    H. Liu, Y . Zhang, H. Zhang, C. Fan, S. Kwong, C.-C. J. Kuo, and X. Fan, “Deep learning-based picture-wise just noticeable distortion prediction model for image compression,” IEEE Trans. Image Process. , vol. 29, pp. 641–656, 2020

  8. [16]

    Deep learning based just noticeable difference and perceptual quality prediction models for compressed video,

    Y . Zhang, H. Liu, Y . Yang, X. Fan, S. Kwong, and C. C. J. Kuo, “Deep learning based just noticeable difference and perceptual quality prediction models for compressed video,” IEEE Trans. Circuit Syst. Video Technol., vol. 32, no. 3, pp. 1197–1212, 2022

  9. [17]

    Lightweight multitask learning for robust jnd prediction using latent space and reconstructed frames,

    S. Nami, F. Pakdaman, M. R. Hashemi, S. Shirmohammadi, and M. Gabbouj, “Lightweight multitask learning for robust jnd prediction using latent space and reconstructed frames,” IEEE Trans. Circuit Syst. Video Technol., vol. 34, no. 9, pp. 8657–8671, 2024

  10. [18]

    A survey on perceptually optimized video coding,

    Y . Zhang, L. Zhu, G. Jiang, S. Kwong, and C.-C. J. Kuo, “A survey on perceptually optimized video coding,” ACM Comput. Surveys , vol. 55, no. 12, pp. 1–37, 2023

  11. [19]

    Perceptual video coding with block- level staircase just noticeable distortion,

    X. Zhang, H. Wang, and T. Tian, “Perceptual video coding with block- level staircase just noticeable distortion,” in Proc. IEEE Int. Conf. Image Process., 2019, pp. 4140–4144

  12. [20]

    Just noticeable distortion-based perceptual rate control in hevc,

    M. Zhou, X. Wei, S. Kwong, W. Jia, and B. Fang, “Just noticeable distortion-based perceptual rate control in hevc,” IEEE Trans. Image Process., vol. 29, pp. 7603–7614, 2020

  13. [21]

    Jnd-based perceptual rate distortion optimization for av1 encoder,

    C. Zhu, L. Song, R. Xie, J. Han, and Y . Xu, “Jnd-based perceptual rate distortion optimization for av1 encoder,” in PCS, 2019, pp. 1–5

  14. [22]

    Jnd-based perceptual optimization for learned image compression,

    F. Ding, J. Jin, L. Meng, and W. Lin, “Jnd-based perceptual optimization for learned image compression,” arXiv preprint arXiv:2302.13092, 2023

  15. [23]

    Videoset: A large-scale compressed video quality dataset based on jnd measurement,

    H. Wang, I. Katsavounidis, J. Zhou, J. Park, S. Lei, X. Zhou, M.-O. Pun, X. Jin, R. Wang, X. Wang et al., “Videoset: A large-scale compressed video quality dataset based on jnd measurement,” J. Vis. Commun. Image Represent., vol. 46, pp. 292–302, 2017

  16. [24]

    Perceptual learned image compression via end-to-end jnd-based optimization,

    F. Pakdaman, S. Nami, and M. Gabbouj, “Perceptual learned image compression via end-to-end jnd-based optimization,” arXiv preprint arXiv:2402.02836, 2024

  17. [25]

    Image coding for machines: an end-to-end learned approach,

    N. Le, H. Zhang, F. Cricri, R. Ghaznavi-Youvalari, and E. Rahtu, “Image coding for machines: an end-to-end learned approach,” in Proc. IEEE Int. Conf. Acoust. Speech Signal Process. , 2021, pp. 1590–1594

  18. [26]

    Towards task-generic image compres- sion: A study of semantics-oriented metrics,

    C. Gao, D. Liu, L. Li, and F. Wu, “Towards task-generic image compres- sion: A study of semantics-oriented metrics,” IEEE Trans. Multimedia , IEEE TRANSACTIONS ON MULTIMEDIA 12 vol. 25, pp. 721–735, 2023

  19. [27]

    Scalable image coding for humans and machines,

    H. Choi and I. V . Baji ´c, “Scalable image coding for humans and machines,” IEEE Trans. Image Process., vol. 31, pp. 2739–2754, 2022

  20. [28]

    Transtic: Transferring transformer-based image compression from human perception to machine perception,

    Y .-H. Chen, Y .-C. Weng, C.-H. Kao, C. Chien, W.-C. Chiu, and W.- H. Peng, “Transtic: Transferring transformer-based image compression from human perception to machine perception,” in Proc. Int. Conf. Comput. Vis., 2023, pp. 23 240–23 250

  21. [29]

    Image coding for machines with omnipotent feature learning,

    R. Feng, X. Jin, Z. Guo, R. Feng, Y . Gao, T. He, Z. Zhang, S. Sun, and Z. Chen, “Image coding for machines with omnipotent feature learning,” in Proc. Eur. Conf. Comput. Vis. , 2022, pp. 510–528

  22. [30]

    Spatial-frequency channels, shape bias, and adversarial robustness,

    A. Subramanian, E. Sizikova, N. Majaj, and D. Pelli, “Spatial-frequency channels, shape bias, and adversarial robustness,” in Proc. Adv. Neural Inf. Process. Syst. , 2023, pp. 4137–4149

  23. [31]

    On human visual contrast sensitivity and machine vision robustness: A comparative study,

    M.-C. Chiu, Y . Wang, D. E. G. Kim, P.-Y . Chen, and X. Ma, “On human visual contrast sensitivity and machine vision robustness: A comparative study,” arXiv preprint arXiv:2212.08650 , 2022

  24. [32]

    Partial success in closing the gap between human and machine vision,

    R. Geirhos, K. Narayanappa, B. Mitzkus, T. Thieringer, M. Bethge, F. A. Wichmann, and W. Brendel, “Partial success in closing the gap between human and machine vision,” in Proc. Adv. Neural Inf. Process. Syst. , 2021, pp. 23 885–23 899

  25. [33]

    Just recognizable distortion for machine vision oriented image and video coding,

    Q. Zhang, S. Wang, X. Zhang, S. Ma, and W. Gao, “Just recognizable distortion for machine vision oriented image and video coding,” Int. J. Comput. Vis., vol. 129, no. 10, pp. 2889–2906, 2021

  26. [34]

    Just noticeable difference for deep machine vision,

    J. Jin, X. Zhang, X. Fu, H. Zhang, W. Lin, J. Lou, and Y . Zhao, “Just noticeable difference for deep machine vision,” IEEE Trans. Circuit Syst. Video Technol., vol. 32, no. 6, pp. 3452–3461, 2022

  27. [35]

    Just noticeable difference for machine perception and generation of regularized adversarial images with minimal perturbation,

    A. K. Akan, E. Akbas, and F. T. Y . Vural, “Just noticeable difference for machine perception and generation of regularized adversarial images with minimal perturbation,” Signal Image Video Process., vol. 16, no. 6, pp. 1595–1606, 2022

  28. [36]

    Perceptual video coding for machines via satisfied machine ratio modeling,

    Q. Zhang, S. Wang, X. Zhang, C. Jia, Z. Wang, S. Ma, and W. Gao, “Perceptual video coding for machines via satisfied machine ratio modeling,” IEEE Trans. Pattern Anal. Mach. Intell. , pp. 1–18, 2024

  29. [37]

    Learning to predict object-wise just recognizable distortion for image and video compression,

    Y . Zhang, H. Lin, J. Sun, L. Zhu, and S. Kwong, “Learning to predict object-wise just recognizable distortion for image and video compression,” IEEE Trans. Multimedia , vol. 26, pp. 5925–5938, 2024

  30. [38]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Transformers for image recognition at scale,” in Proc. Int. Conf. Learn. Represent., 2021

  31. [39]

    Microsoft coco: Common objects in context,

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll ´ar, and C. L. Zitnick, “Microsoft coco: Common objects in context,” in Proc. Eur. Conf. Comput. Vis. , 2014, pp. 740–755

  32. [40]

    Big transfer (bit): General visual representation learning,

    A. Kolesnikov, L. Beyer, X. Zhai, J. Puigcerver, J. Yung, S. Gelly, and N. Houlsby, “Big transfer (bit): General visual representation learning,” in Proc. Eur. Conf. Comput. Vis. , 2020, p. 491–507

  33. [41]

    Fixing the train-test resolution discrepancy,

    H. Touvron, A. Vedaldi, M. Douze, and H. J ´egou, “Fixing the train-test resolution discrepancy,” in Proc. Adv. Neural Inf. Process. Syst. , 2019

  34. [42]

    An empirical study of training self- supervised vision transformers,

    X. Chen, S. Xie, and K. He, “An empirical study of training self- supervised vision transformers,” in Proc. Int. Conf. Comput. Vis. , 2021, pp. 9620–9629

  35. [43]

    Rethinking the inception architecture for computer vision,

    C. Szegedy, V . Vanhoucke, S. Ioffe, J. Shlens, and Z. Wojna, “Rethinking the inception architecture for computer vision,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. , 2016, pp. 2818–2826

  36. [44]

    Image quality assess- ment: from error visibility to structural similarity,

    Z. Wang, A. Bovik, H. Sheikh, and E. Simoncelli, “Image quality assess- ment: from error visibility to structural similarity,” IEEE Transactions on Image Processing , vol. 13, no. 4, pp. 600–612, 2004

  37. [45]

    The unreasonable effectiveness of deep features as a perceptual metric,

    R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang, “The unreasonable effectiveness of deep features as a perceptual metric,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 586–595

  38. [46]

    Object- wise just recognizable distortion dataset,

    Y . Zhang, H. Lin, J. Sun, L. Zhu, and S. Kwong, “Object- wise just recognizable distortion dataset,” 2023. [Online]. Available: https://dx.doi.org/10.21227/q5fq-d638

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.