pith. sign in

arxiv: 2606.26285 · v1 · pith:BFKIRQTNnew · submitted 2026-06-24 · 💻 cs.CR · cs.AI

TEMPO-Diffusion: Temporally Exposed Malicious Poisoning of Diffusion Models

William Aiken , Paula Branco , Guy-Vincent Jourdan , Iosif-Viorel Onut This is my paper

Pith reviewed 2026-06-26 01:27 UTC · model grok-4.3

classification 💻 cs.CR cs.AI
keywords backdoor attacksdiffusion modelsdata poisoningsynthetic datatargeted attacksTEMPO-Diffusiontraffic sign recognition
0
0 comments X

The pith

TEMPO-Diffusion uses time-conditioned triggers to localize backdoors in diffusion models, poisoning class-specific synthetic images that then compromise downstream classifiers.

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

The paper shows how to tie malicious behavior in diffusion models to specific moments during image generation so that the shift stays inside the normal data distribution. This lets an attacker target particular output classes, embed multiple hidden features at chosen image locations, and perform in-painting attacks. Experiments on CIFAR10, GTSRB, and the new CALISA traffic-sign dataset demonstrate that classifiers trained on the resulting synthetic data exhibit high attack success rates. A reader should care because diffusion models are now common sources of training data, so such poisoning could affect any system that relies on their output without obvious signs of tampering.

Core claim

TEMPO-Diffusion is a targeted backdoor framework that localizes the malicious distribution shift to a temporal, in-distribution exposure. It supports attacks on and to specific classes, multiple sub-image backdoors that reconstruct chosen features at chosen locations, and in-painting with time-conditioned triggers. Across CIFAR10, GTSRB, and CALISA the method reliably poisons class-specific synthetic data generation and produces high attack success rates in downstream classifiers trained on that data.

What carries the argument

Time-conditioned triggers that localize the malicious distribution shift to specific temporal exposures while remaining in-distribution.

If this is right

  • Targeted attacks become possible on and to chosen output classes in the generated images.
  • Multiple independent backdoors can be placed inside different regions of the same or different output images.
  • In-painting tasks can be poisoned using the same time-conditioned mechanism.
  • Classifiers trained on the poisoned synthetic data show high attack success rates on the tested datasets.

Where Pith is reading between the lines

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

  • Detection methods for diffusion-model poisoning may need to inspect generation timing or step-wise distribution statistics rather than final images alone.
  • Organizations that use diffusion-generated data for training should add integrity checks on the generative model itself.
  • The same temporal-trigger idea could be tested on other iterative generative processes that have an explicit time or step parameter.
  • In domains such as autonomous driving, pipelines that rely on synthetic traffic-sign data would become a new attack surface.

Load-bearing premise

Time-conditioned triggers can localize the malicious distribution shift to specific temporal exposures while remaining in-distribution and undetectable during normal generation.

What would settle it

Generate images both with and without the chosen time trigger and measure whether the downstream classifier attack success rate falls to chance level on the no-trigger images.

Figures

Figures reproduced from arXiv: 2606.26285 by Guy-Vincent Jourdan, Iosif-Viorel Onut, Paula Branco, William Aiken.

Figure 1
Figure 1. Figure 1: Threat model for TEMPO-Diffusion. Despite these advancements, the security implications of diffusion-based gen￾erative models remain relatively under-explored. Backdoor attacks against diffu￾sion models have primarily focused on unrealistic threat models, where attack￾ers are assumed to have control over the inference-time noise seed in order to generate the desired output. In particular, existing threat m… view at source ↗
Figure 2
Figure 2. Figure 2: Forward diffusion process for abridged, hann, and box [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Unbalanced vs. balanced GTSRB backdoor and clean noise rates. Rates [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Backdoor and clean noise rates across TSE intervals and TWT functions. [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Confusion matrices for the ResNet models trained on DDIM samples. [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
read the original abstract

Noise-based backdoor attacks on diffusion models typically rely on input-time trigger injection, untargeted activation, and out-of-distribution target generation. Such assumptions reduce both the stealthiness and the practical relevance of these attacks. In this work, we present TEMPO-Diffusion, a targeted backdoor framework that localizes the malicious distribution shift to a temporal, in-distribution exposure. TEMPO-Diffusion supports: (i) targeted attacks on and to specific classes, (ii) multiple sub-image backdoors that reconstruct specific features within multiple, different output images and at multiple locations, and (iii) in-painting with time-conditioned triggers. To study relevant, practical security concerns in leveraging backdoored diffusion models for synthetic training data, we also introduce CALISA: a balanced, region-aware traffic-sign dataset emphasizing Canadian and U.S. road signs. Across CIFAR10, GTSRB, and CALISA, our experiments show that TEMPO-Diffusion can reliably poison class-specific synthetic data generation and induce high attack success rates in downstream classifiers trained on that data.

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

1 major / 0 minor

Summary. The paper proposes TEMPO-Diffusion, a targeted backdoor framework for diffusion models that localizes malicious distribution shifts to temporal, in-distribution exposures via time-conditioned triggers. It claims support for targeted class-specific attacks, multiple sub-image backdoors reconstructing specific features at multiple locations, and in-painting. The work introduces the CALISA dataset (balanced, region-aware traffic signs) and reports that experiments on CIFAR10, GTSRB, and CALISA show reliable poisoning of class-specific synthetic data with high attack success rates in downstream classifiers trained on the poisoned outputs.

Significance. If the empirical claims hold with proper controls and metrics, the work would highlight a practical security risk in using diffusion models to generate synthetic training data for classifiers, extending backdoor concerns to temporally localized, in-distribution triggers. The introduction of CALISA provides a new benchmark for region-aware traffic-sign tasks.

major comments (1)
  1. [Abstract] Abstract: The central claims of reliable poisoning and high attack success rates across CIFAR10, GTSRB, and CALISA are asserted without any experimental details, metrics (e.g., ASR values, clean accuracy), baselines, controls, or ablation results on temporal localization; this prevents evaluation of whether the time-conditioned triggers remain in-distribution and undetectable.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their review. We address the single major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claims of reliable poisoning and high attack success rates across CIFAR10, GTSRB, and CALISA are asserted without any experimental details, metrics (e.g., ASR values, clean accuracy), baselines, controls, or ablation results on temporal localization; this prevents evaluation of whether the time-conditioned triggers remain in-distribution and undetectable.

    Authors: We acknowledge that the provided abstract is concise and does not include specific numerical results, baselines, or references to ablations. In the revised version we will expand the abstract to include key metrics (ASR and clean accuracy values from the experiments), a brief note on the controls, and an explicit reference to the temporal-localization ablations. The full experimental details, baselines, metrics, and ablations demonstrating that the time-conditioned triggers remain in-distribution are already contained in Sections 4 and 5; the abstract revision will simply surface these results for readers evaluating the claims. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

This is an empirical security/attack paper describing a new backdoor framework (TEMPO-Diffusion) for diffusion models, with targeted class attacks, multi-feature backdoors, time-conditioned in-painting, and downstream ASR evaluation on CIFAR10/GTSRB/CALISA. The provided text contains no mathematical derivations, no equations, no fitted parameters renamed as predictions, and no self-citation chains. Central claims rest on experimental results rather than any derivation that reduces to its own inputs by construction. This matches the default expectation for non-circular empirical work.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no information on free parameters, axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5732 in / 1098 out tokens · 20096 ms · 2026-06-26T01:27:42.603963+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

40 extracted references · 17 canonical work pages

  1. [1]

    In: 2024 21st Annual International Conference on Privacy, Security and Trust (PST)

    Aiken, W., Branco, P., Jourdan, G.V.: DevilDiffusion: Embedding hidden noise backdoors into diffusion models. In: 2024 21st Annual International Conference on Privacy, Security and Trust (PST). pp. 1–11 (2024).https://doi.org/10.1109/ PST62714.2024.10788056

    arXiv 2024

  2. [2]

    In: Advances and Trends in Artificial Intelligence

    Aiken, W., Branco, P., Jourdan, G.V.: Lost in the noise: Evading and de- tecting backdoors in conditional diffusion models. In: Advances and Trends in Artificial Intelligence. Theory and Applications. pp. 481–493. Springer Na- ture Singapore, Singapore (2026).https://doi.org/https://doi.org/10.1007/ 978-981-96-8889-0_41

    2026

  3. [3]

    Phase-informed bayesian ensemble models improve performance of covid-19 forecasts,

    An, S., Chou, S.Y., Zhang, K., Xu, Q., Tao, G., Shen, G., Cheng, S., Ma, S., Chen,P.Y.,Ho,T.Y.,Zhang,X.:Elijah:Eliminatingbackdoorsinjectedindiffusion models via distribution shift. Proceedings of the AAAI Conference on Artificial TEMPO-Diffusion 17 Intelligence38(10), 10847–10855 (Mar 2024).https://doi.org/10.1609/aaai. v38i10.28958

    work page doi:10.1609/aaai 2024

  4. [4]

    Transactions on Machine Learn- ing Research (2023),https://openreview.net/forum?id=DlRsoxjyPm

    Azizi, S., Kornblith, S., Saharia, C., Norouzi, M., Fleet, D.J.: Synthetic data from diffusion models improves imagenet classification. Transactions on Machine Learn- ing Research (2023),https://openreview.net/forum?id=DlRsoxjyPm

    2023

  5. [5]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

    Bhosale, M., Wasi, A., Zhai, Y., Tian, Y., Border, S., Xi, N., Sarder, P., Yuan, J., Doermann, D., Gong, X.: PathDiff: Histopathology image synthesis with un- paired text and mask conditions. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 22415–22424 (October 2025)

    2025

  6. [6]

    In: SAE Technical Paper 2023-01-7046

    Bie, X., Zhang, S., Meng, C., Mei, J., Li, J., He, X.: Synthetic data for 2D road marking detection in autonomous driving. In: SAE Technical Paper 2023-01-7046. pp. 1–10. SAE International (2023).https://doi.org/10.4271/2023-01-7046, presented at the SAE 2023 Intelligent and Connected Vehicles Symposium

    work page doi:10.4271/2023-01-7046 2023

  7. [7]

    Master’s thesis, Linköping University, Computer Vision (2023)

    Carlson, J., Byman, L.: Generation of Synthetic Traffic Sign Images using Diffusion Models. Master’s thesis, Linköping University, Computer Vision (2023)

    2023

  8. [8]

    In: 2023 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR)

    Chen, W., Song, D., Li, B.: TrojDiff: Trojan attacks on diffusion models with diverse targets. In: 2023 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR). pp. 4035–4044 (2023).https://doi.org/10.1109/ CVPR52729.2023.00393

    arXiv 2023

  9. [9]

    Chen, X., Liu, C., Li, B., Lu, K., Song, D.: Targeted backdoor attacks on deep learning systems using data poisoning (2017), arXiv preprint arXiv:1712.05526

    Pith/arXiv arXiv 2017

  10. [10]

    Chou, S.Y., Chen, P.Y., Ho, T.Y.: How to backdoor diffusion models? In: 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 4015–4024 (2023).https://doi.org/10.1109/CVPR52729.2023.00391

    work page doi:10.1109/cvpr52729.2023.00391 2023

  11. [11]

    Advances in Neural Information Processing Sys- tems36(2024), available:https://papers.nips.cc/paper_files/paper/2023/ hash/6b055b95d689b1f704d8f92191cdb788-Abstract-Conference.html

    Chou, S.Y., Chen, P.Y., Ho, T.Y.: VillanDiffusion: A unified backdoor attack framework for diffusion models. Advances in Neural Information Processing Sys- tems36(2024), available:https://papers.nips.cc/paper_files/paper/2023/ hash/6b055b95d689b1f704d8f92191cdb788-Abstract-Conference.html

    2024

  12. [12]

    Proceedings of the AAAI Conference on Artificial Intelli- gence39(11), 11185–11193 (Apr 2025).https://doi.org/10.1609/aaai.v39i11

    Dai,W.,Lu,L.,Li,Z.:Diffusion-basedsyntheticdatagenerationforvisible-infrared person re-identification. Proceedings of the AAAI Conference on Artificial Intelli- gence39(11), 11185–11193 (Apr 2025).https://doi.org/10.1609/aaai.v39i11. 33216

    work page doi:10.1609/aaai.v39i11 2025

  13. [13]

    Vietnam Journal of Computer Science09(03), 333–348 (2022),https://doi.org/10.1142/S2196888822500191

    Dewi, C., Chen, R.C., Liu, Y.T.: Synthetic traffic sign image generation applying generative adversarial networks. Vietnam Journal of Computer Science09(03), 333–348 (2022),https://doi.org/10.1142/S2196888822500191

    work page doi:10.1142/s2196888822500191 2022

  14. [14]

    In: Computer Vision – ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXIII

    Ertler, C., Mislej, J., Ollmann, T., Porzi, L., Neuhold, G., Kuang, Y.: The Map- illary traffic sign dataset for detection and classification on a global scale. In: Computer Vision – ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXIII. p. 68–84. Springer-Verlag, Berlin, Heidel- berg (2020).https://doi.org/10.1007/...

    work page doi:10.1007/978-3-030-58592-1_5 2020

  15. [15]

    In: Pro- ceedings of the 37th International Conference on Neural Information Processing Systems

    Fang, G., Ma, X., Wang, X.: Structural pruning for diffusion models. In: Pro- ceedings of the 37th International Conference on Neural Information Processing Systems. NIPS ’23, Curran Associates Inc., Red Hook, NY, USA (2023)

    2023

  16. [16]

    arXiv preprint arXiv:2303.18037 (2023)

    Ge, J.: Traffic sign recognition dataset and data augmentation. arXiv preprint arXiv:2303.18037 (2023)

    arXiv 2023

  17. [17]

    IEEE Access7, 47230–47244 (2019).https: //doi.org/10.1109/ACCESS.2019.2909068

    Gu, T., Liu, K., Dolan-Gavitt, B., Garg, S.: BadNets: Evaluating backdooring attacks on deep neural networks. IEEE Access7, 47230–47244 (2019).https: //doi.org/10.1109/ACCESS.2019.2909068

    work page doi:10.1109/access.2019.2909068 2019

  18. [18]

    In: Proceedings of the Thirty- Ninth AAAI Conference on Artificial Intelligence and Thirty-Seventh Conference 18 W

    Guan, Z., Hu, M., Li, S., Vullikanti, A.K.: Ufid: a unified framework for black-box input-level backdoor detection on diffusion models. In: Proceedings of the Thirty- Ninth AAAI Conference on Artificial Intelligence and Thirty-Seventh Conference 18 W. Aiken et al. on Innovative Applications of Artificial Intelligence and Fifteenth Symposium on Educational...

    work page doi:10.1609/aaai.v39i26.34941 2025

  19. [19]

    doi:10.1109/CVPR52734.2025.00499

    Han, Y., Zhao, B., Chu, R., Luo, F., Sikdar, B., Lao, Y.: UIBDiffusion: Univer- sal imperceptible backdoor attack for diffusion models. In: 2025 IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR). pp. 19186–19196 (2025).https://doi.org/10.1109/CVPR52734.2025.01787

    work page doi:10.1109/cvpr52734.2025.01787 2025

  20. [20]

    He,K.,Zhang,X.,Ren,S.,Sun,J.:Deepresiduallearningforimagerecognition.In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (June 2016)

    2016

  21. [21]

    https: //doi.org/10.5555/3495724.3496298

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Ad- vances in neural information processing systems33, 6840–6851 (2020), doi: 10.5555/3495724.3496298

    work page doi:10.5555/3495724.3496298 2020

  22. [22]

    In: 2025 IEEE International Con- ference on Multimedia and Expo (ICME)

    Jiang, H., Xiao, J., Hu, X., Chen, T., Zhao, J.: Diff-Cleanse: Identifying and mit- igating backdoor attacks in diffusion models. In: 2025 IEEE International Con- ference on Multimedia and Expo (ICME). pp. 1–6 (2025).https://doi.org/10. 1109/ICME59968.2025.11210014

    arXiv 2025

  23. [23]

    Technical Report (2009), available:https://www.cs.toronto.edu/~kriz/ cifar.html

    Krizhevsky, A., Hinton, G., et al.: Learning multiple layers of features from tiny images. Technical Report (2009), available:https://www.cs.toronto.edu/~kriz/ cifar.html

    2009

  24. [24]

    Nature Medicine30(4), 1166–1173 (2024).https://doi.org/10.1038/s41591-024-02838-6

    Ktena, I., Wiles, O., Albuquerque, I., Rebuffi, S.A., Tanno, R., Roy, A.G., Azizi, S., Belgrave, D., Kohli, P., Cemgil, T., et al.: Generative models improve fairness of medical classifiers under distribution shifts. Nature Medicine30(4), 1166–1173 (2024).https://doi.org/10.1038/s41591-024-02838-6

    work page doi:10.1038/s41591-024-02838-6 2024

  25. [25]

    Li, S., Ma, J., Cheng, M.: Learnable invisible backdoor for diffusion models (2024), available:https://openreview.net/forum?id=scFfMOOGD8

    2024

  26. [26]

    In: Proceedings of the 36th International Conference on Neural Information Processing Systems

    Lu, C., Zhou, Y., Bao, F., Chen, J., Li, C., Zhu, J.: DPM-solver: a fast ODE solver for diffusion probabilistic model sampling in around 10 steps. In: Proceedings of the 36th International Conference on Neural Information Processing Systems. NIPS ’22, Curran Associates Inc., Red Hook, NY, USA (2022),https://dl.acm.org/ doi/10.5555/3600270.3600688

    work page doi:10.5555/3600270.3600688 2022

  27. [27]

    Machine Intelligence Research 22(4), 730–751 (2025).https://doi.org/10.1007/s11633-025-1562-4

    Lu, C., Zhou, Y., Bao, F., Chen, J., Li, C., Zhu, J.: DPM-Solver++: Fast solver for guided sampling of diffusion probabilistic models. Machine Intelligence Research 22(4), 730–751 (2025).https://doi.org/10.1007/s11633-025-1562-4

    work page doi:10.1007/s11633-025-1562-4 2025

  28. [28]

    In: International Conference on Learning Representations (2018),https://openreview.net/forum?id=rJzIBfZAb

    Madry, A., Makelov, A., Schmidt, L., Tsipras, D., Vladu, A.: Towards deep learning models resistant to adversarial attacks. In: International Conference on Learning Representations (2018),https://openreview.net/forum?id=rJzIBfZAb

    2018

  29. [29]

    IEEE Transactions on Intelligent Transportation Systems13(4), 1484–1497 (2012)

    Mogelmose, A., Trivedi, M.M., Moeslund, T.B.: Vision-based traffic sign detec- tion and analysis for intelligent driver assistance systems: Perspectives and survey. IEEE Transactions on Intelligent Transportation Systems13(4), 1484–1497 (2012). https://doi.org/10.1109/TITS.2012.2209421

    work page doi:10.1109/tits.2012.2209421 2012

  30. [30]

    Rahimi, P., Teney, D., Marcel, S.: AugGen: Synthetic augmentation using diffusion modelscanimproverecognition.In:TheThirty-ninthAnnualConferenceonNeural Information Processing Systems (2025)

    2025

  31. [31]

    Computer Optics40(2), 294–300 (2016).https://doi.org/10.18287/ 2412-6179-2016-40-2-294-300

    Shakhuro, V.I., Konushin, A.S.: Russian traffic sign images dataset. Computer Optics40(2), 294–300 (2016).https://doi.org/10.18287/ 2412-6179-2016-40-2-294-300

    2016

  32. [32]

    Song,J.,Meng,C.,Ermon,S.:Denoisingdiffusionimplicitmodels.In:International Conferenceon LearningRepresentations(2021),https://openreview.net/forum? id=St1giarCHLP TEMPO-Diffusion 19

    2021

  33. [33]

    doi: 10.1016/j.neunet.2012.02.016

    Stallkamp, J., Schlipsing, M., Salmen, J., Igel, C.: Man vs. computer: Bench- marking machine learning algorithms for traffic sign recognition. Neural Networks (2012), doi: 10.1016/j.neunet.2012.02.016

    work page doi:10.1016/j.neunet.2012.02.016 2012

  34. [34]

    Transactions on Machine Learning Research (2025),https://openreview.net/ forum?id=SfqCaAOF1S

    Sui, Y., Phan, H., Xiao, J., Zhang, T., Tang, Z., Shi, C., Wang, Y., Chen, Y., Yuan, B.: DisDet: Exploring detectability of backdoor attack on diffusion models. Transactions on Machine Learning Research (2025),https://openreview.net/ forum?id=SfqCaAOF1S

    2025

  35. [35]

    In: 2019 International Joint Conference on Neural Networks (IJCNN)

    Torres, L.T., Paixão, T.M., Berriel, R.F., De Souza, A.F., Badue, C., Sebe, N., Oliveira-Santos, T.: Effortless deep training for traffic sign detection using tem- plates and arbitrary natural images. In: 2019 International Joint Conference on Neural Networks (IJCNN). pp. 1–7 (2019).https://doi.org/10.1109/IJCNN. 2019.8852086

    work page doi:10.1109/ijcnn 2019

  36. [36]

    arXiv preprint arXiv.1912.02771 (2019)

    Turner, A., Tsipras, D., Madry, A.: Label-consistent backdoor attacks. arXiv preprint arXiv.1912.02771 (2019)

    arXiv 1912

  37. [37]

    In: 2024 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

    Valvano, G., Agostino, A., De Magistris, G., Graziano, A., Veneri, G.: Control- lable image synthesis of industrial data using stable diffusion. In: 2024 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). pp. 5342–5351 (2024).https://doi.org/10.1109/WACV57701.2024.00527

    work page doi:10.1109/wacv57701.2024.00527 2024

  38. [38]

    arXiv preprint arXiv:2204.13902 (2022)

    Zhang, Q., Chen, Y.: Fast sampling of diffusion models with exponential integrator. arXiv preprint arXiv:2204.13902 (2022)

    arXiv 2022

  39. [39]

    In: Oh, A., Naumann, T., Globerson, A., Saenko, K., Hardt, M., Levine, S

    Zhao, W., Bai, L., Rao, Y., Zhou, J., Lu, J.: UniPC: A unified predictor- corrector framework for fast sampling of diffusion models. In: Oh, A., Naumann, T., Globerson, A., Saenko, K., Hardt, M., Levine, S. (eds.) Advances in Neu- ral Information Processing Systems. vol. 36, pp. 49842–49869. Curran Associates, Inc. (2023),https://proceedings.neurips.cc/pa...

    2023

  40. [40]

    In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

    Zhu, Z., Liang, D., Zhang, S., Huang, X., Li, B., Hu, S.: Traffic-sign detection and classification in the wild. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). pp. 2110–2118 (2016).https://doi.org/10.1109/ CVPR.2016.232 A CALISA Dataset Details Despite the abundance of large vision datasets, datasets for the detection and clas...

    2016