Conference paper · 2022 · Open access

Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate

Mi Lu, Hao Wang, Yonglong Tian, Hao He, Nir Shavit

Proceedings of the AAAI Conference on Artificial Intelligence, 36(9), 10042–10050 · Published 28 June 2022 · DOI 10.1609/aaai.v36i9.21243

Open PDFView at the publisherCite this paper

In brief

“Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate” is a 2022 conference paper by Mi Lu and 4 co-authors, published in Proceedings of the AAAI Conference on Artificial Intelligence. OpenAlex counts 51 citations as of 22 November 2026, which puts it in the top 4% of works published in 2022 by citations. It is open access: the full text is free to read at Proceedings of the AAAI Conference on Artificial Intelligence.

Citations
51
Most cited in
2025 (13)
Field-weighted impact
2.15
References
60
Authors
5

Abstract

Uncertainty estimation is an essential step in the evaluation of the robustness for deep learning models in computer vision, especially when applied in risk-sensitive areas. However, most state-of-the-art deep learning models either fail to obtain uncertainty estimation or need significant modification (e.g., formulating a proper Bayesian treatment) to obtain it. Most previous methods are not able to take an arbitrary model off the shelf and generate uncertainty estimation without retraining or redesigning it. To address this gap, we perform a systematic exploration into training-free uncertainty estimation for dense regression, an unrecognized yet important problem, and provide a theoretical construction justifying such estimations. We propose three simple and scalable methods to analyze the variance of outputs from a trained network under tolerable perturbations: infer-transformation, infer-noise, and infer-dropout. They operate solely during the inference, without the need to re-train, re-design, or fine-tune the models, as typically required by state-of-the-art uncertainty estimation methods. Surprisingly, even without involving such perturbations in training, our methods produce comparable or even better uncertainty estimation when compared to training-required state-of-the-art methods. Code is available at https://github.com/lumi9587/train-free-uncertainty.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Lu, M., Wang, H., Tian, Y., He, H., & Shavit, N. (2022). Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate. Proceedings of the AAAI Conference on Artificial Intelligence, 36(9), 10042–10050. https://doi.org/10.1609/aaai.v36i9.21243

MLA 9

Lu, Mi, et al. “Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate.” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 36, no. 9, 2022, pp. 10042–10050. https://doi.org/10.1609/aaai.v36i9.21243.

Chicago

Lu, Mi, Hao Wang, Yonglong Tian, Hao He, and Nir Shavit. 2022. “Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate.” Proceedings of the AAAI Conference on Artificial Intelligence 36 (9): 10042–10050. https://doi.org/10.1609/aaai.v36i9.21243.

Harvard

Lu, M. et al. (2022) ‘Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate’, Proceedings of the AAAI Conference on Artificial Intelligence, 36(9), pp. 10042–10050. doi:10.1609/aaai.v36i9.21243.

IEEE

M. Lu, H. Wang, Y. Tian, H. He, and N. Shavit, “Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate,” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 36, no. 9, pp. 10042–10050, 2022, doi: 10.1609/aaai.v36i9.21243.

BibTeX

@inproceedings{lu2022training,
  title     = {{Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate}},
  author    = {Lu, Mi and Wang, Hao and Tian, Yonglong and He, Hao and Shavit, Nir},
  booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year      = {2022},
  volume    = {36},
  number    = {9},
  pages     = {10042--10050},
  doi       = {10.1609/aaai.v36i9.21243},
  url       = {https://doi.org/10.1609/aaai.v36i9.21243}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Citations by year

YearCitationsShare of the best year
202044
202144
20221010
202388
202455
20251313
202677

Its field-weighted citation impact is 2.15: it has been cited 2.15 times as often as the average work of the same field, type and year (1.00 is the world average).

Topic and keywords

Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications

  • uncertainty estimation
  • training-free methods
  • sensitivity analysis
  • surrogate models
  • output variance
  • perturbation analysis
  • deep neural network robustness
  • computer vision

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

Lu, M., Wang, H., Tian, Y., He, H., & Shavit, N. (2022). Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate. Proceedings of the AAAI Conference on Artificial Intelligence, 36(9), 10042–10050. https://doi.org/10.1609/aaai.v36i9.21243

How many times has this paper been cited?

51 times, according to OpenAlex on 22 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at Proceedings of the AAAI Conference on Artificial Intelligence: sign in and open the PDF from this page.

Who wrote this paper?

Mi Lu, Hao Wang, Yonglong Tian, Hao He, Nir Shavit.

What is the DOI of this paper?

10.1609/aaai.v36i9.21243, permanently at https://doi.org/10.1609/aaai.v36i9.21243. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 22 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.