Pith. sign in

REVIEW 4 major objections 4 minor 36 references

An In-Depth Examination of Risk Assessment in Multi-Class Classification Algorithms

T0 review · 4 major / 4 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 Inverse conformal prediction gives a conservative, model-agnostic estimate of a classifier's misclassification probability, with no retraining or hyperparameter tuning beyond the calibration set size.

desk verdict InvCP is a sensible extension of inverse conformal prediction to classification with a broad empirical study, but the paper's key safety claim—conservative risk estimates—fails on a simple exchangeable construction, so treat it as empirical, not guaranteed. read the letter →

arxiv 2412.04166 v1 pith:KYR2H4ER submitted 2024-12-05 cs.LG cs.NAmath.NA

Disha Ghandwani , Neeraj Sarna , Yuanyuan Li , Yang Lin This is my paper · ORCID
classification cs.LGcs.NAmath.NA
keywords riskassessmentcalibrationuncertaintyquantificationAIsafetyconformalpredictioninversemisclassificationprobabilitymulti-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 asks a practical safety question: how likely is a trained classifier to be wrong on the next input? It proposes a new method, InvCP, that inverts conformal prediction: instead of building a prediction interval with a target confidence, it treats the model's own output as the interval and finds the smallest conformal interval that contains it, then uses that interval's confidence level as the estimated probability that the model is correct. Averaging this over a test set yields an estimate of the misclassification probability. In experiments across five image datasets and many model architectures, InvCP is often the most accurate risk-assessment method for datasets with fewer labels, is always conservative (it never underestimates risk in the tested cases), and requires no optimization or bin-count choices, unlike calibration baselines.

What carries the argument

The central object is the InvCP algorithm built on the Adaptive Prediction Sets (APS) score function. For a calibration set, scores s(x,y) are computed as the cumulative probability from the most likely class down to class y, then sorted. For a test input, the algorithm computes the score of the model's output interval I(X) as the largest APS score among its members, then finds the smallest calibration quantile q̂(X) that is at least that score. The rank γ(X) of q̂(X) among the n calibration scores defines α(X) = 1 - γ(X)/(n+1), and averaging α(X) over the test set estimates the misclassification probability. This inverts the conformal prediction problem: rather than 'given α, find the interval', it solves 'given the interval, find α'.

What would settle it

A concrete test: apply InvCP to a dataset where the calibration and test sets are drawn from different distributions (e.g., a covariate shift) and compute δ = α̂ - αEmp. Any negative δ would falsify the conservativeness claim as stated, since the paper provides no theoretical guarantee to rule this out.

Watch

Extended reading notes

Core claim

We introduce inverse conformal prediction (InvCP) for risk assessment in multi-class classification. Given a model's output interval I(X) (the top-1 class or the top-k set), we compute the smallest conformal prediction interval T(X; α(X)) that contains I(X) using the Adaptive Prediction Sets score, and we read off α(X) as the miss-coverage level of I(X). Averaging α(X) over a held-out test set gives an estimate of the misclassification probability P(Y ≠ Ŷ(X)). Across CIFAR-100, CIFAR-10, Flowers102, ImageNet-V1, and Places365, with tree-based and convolutional models, InvCP outperforms temperature scaling, histogram binning, and isotonic regression for datasets with relatively few labels, remains competitive for large-label datasets, and is conservative in every experiment. We do not claim a theoretical guarantee of conservativeness; that question is left open, as noted in Remark III.1.

Load-bearing premise

The claim that InvCP never underestimates the misclassification probability is an empirical observation on the tested datasets, not a proven property.

Editorial extensions

If this is right

  • For safety-critical applications, InvCP offers a way to estimate a classifier's failure probability without retraining the model or assuming a parametric form of the data distribution.
  • On datasets with relatively few labels (CIFAR-10, CIFAR-100, Flowers102), InvCP generally matches or beats standard calibration techniques in accuracy while remaining conservative.
  • InvCP's only user-chosen parameter is the calibration set size; its performance is stable across calibration sizes, unlike histogram binning and isotonic regression, which require choosing bin counts and degrade with fewer calibration points.
  • Because InvCP is model-agnostic and does not optimize a loss, it can be applied to any classifier that outputs class probabilities, including tree ensembles and deep networks, with minimal implementation effort.
  • The method extends naturally to top-k model outputs, estimating the probability that the true class is not among the model's top k predictions.

Reading between the lines

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

  • If conservativeness could be proven theoretically, InvCP would become a drop-in audit tool that provides certified upper bounds on misclassification risk for any classifier; the paper's Remark III.1 suggests the regression analogue has such a guarantee, making a proof for classification a plausible next step.
  • The experiments focus on image datasets, so a natural testable extension is whether InvCP's conservativeness and accuracy hold on tabular or text data, where score distributions and model overconfidence behave differently.
  • The paper explicitly notes that data drift would break the exchangeability assumption; extending InvCP with the covariate-shift adjustment techniques already developed for the regression risk-assessment line of work would be a direct continuation.
  • The observed instability of histogram binning on large-label datasets suggests a hybrid strategy—use InvCP for small label spaces and calibrated binning for very large ones—which the paper does not itself explore.
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 / 4 minor

Summary. The paper studies the risk-assessment problem for multi-class classifiers: estimating the probability that the model's predicted label (or top-k set) is wrong. It compares standard calibration approaches (softmax, Platt scaling, histogram binning, isotonic regression) with a proposed method called Inverse Conformal Prediction (InvCP), which treats the model's predicted class set as a conformal prediction interval and estimates its miscoverage level from calibration scores. Experiments on CIFAR-10/100, Flowers102, ImageNet, and Places365 across tree-based and CNN architectures report the deviation δ between the estimated and empirical misclassification rates. The paper concludes that InvCP is generally the best for datasets with fewer labels and is conservative throughout, though no method dominates.

Significance. If the central claims held, the paper would offer a simple, model-agnostic procedure for conservative risk estimates in safety-critical applications, which is a genuinely useful goal. The empirical comparison is extensive, spanning five datasets, eleven model architectures, and several calibration baselines. However, the key advertised property of InvCP — conservativeness, δ ≥ 0 — is not guaranteed and is in fact violated by a simple exchangeable counterexample. This undermines the main safety-critical motivation and requires a substantial reframing of the contribution. The experimental survey alone may still be of interest, but the paper's current framing overstates what the method delivers.

major comments (4)
  1. [§III-B, Eq. (14) and §V] The central claim that InvCP provides a conservative risk estimate is not supported and is demonstrably false. In the construction of T(X; α(X)), the conformal coverage guarantee applies to the larger set T, not to I(X) = {Ŷ(X)}. Since I(X) ⊆ T, we have P(Y ∉ I) = P(Y ∉ T) + P(Y ∈ T \ I), and the second term can be large. For the three-class exchangeable construction described in the stress-test note (t ~ U(0.51, 0.99), model outputs (t, 0.01, 0.99−t), true label always class 2), the true error is 1 while the InvCP estimate averages ≈ 0.52, giving δ ≈ −0.48 in an i.i.d. setting. This directly contradicts the conclusion's statement that InvCP 'was conservative throughout.' The authors should either prove a sufficient condition for conservativeness (e.g., a condition on the conditional distribution of scores) or explicitly restrict the claim to the studied datasets and add a prominent caveat that InvCP can underestimate risk.
  2. [§IV, 'Performance Metrics' and Figures 1–3] The experimental comparison reports only averages of δ over 100 random splits, with no standard deviations, confidence intervals, or significance tests. Many of the differences between methods are small (e.g., within 0.02–0.05 in Figure 1a), so it is impossible to assess whether the reported ordering is meaningful or within noise. The authors should report variance or perform pairwise significance tests, especially since the conclusion draws comparative claims such as 'InvCP provides the best results in general.'
  3. [§IV-A, 'Results for different calibration points and bins'] The number of bins M for HIST-BIN and ISO-REG is a critical hyperparameter, yet the main comparisons in Figures 1–3 do not state which M was used. Since Figure 5 shows that performance changes substantially with M, the reported results are not reproducible without this information. Additionally, the extension of these calibration methods to top-k outputs is not described; the paper should clarify whether a one-vs-all approach is used and how the calibrated probability for a set is computed.
  4. [Eq. (10), Eq. (12), and Algorithm 1] There are precision issues in the definition of α(X). First, Eq. (12) defines γ(X) := min{i : s_i ≥ s(X, I(X))}, while Algorithm 1 writes γ(X) ← min{k : s_i* ≤ s_k}; the inequality direction should be made consistent. Second, the arg-min in Eq. (10) over α′ ∈ [0,1] may be ill-defined at α′ = 0, where the conformal quantile is not defined (or is infinite), and ties in the arg-min are not addressed. This should be formalized.
minor comments (4)
  1. [§III-A] The sentence 'Empirically, we found that the LAC score function ... provided inferior results' is not substantiated with a figure or table. If LAC is used as a design choice, the comparison should be shown or the statement removed.
  2. [§IV-A] The paper claims InvCP is 'free of optimization and hyper-parameter tuning,' but the choice of score function and the calibration set size n are still user choices. Qualify this statement.
  3. [Abstract and Introduction] The phrase 'model and data-distribution agnostic' is too strong: conformal prediction requires exchangeability of the calibration and test data, which is a distributional assumption. The authors should state 'agnostic to the model family and to the functional form of the class probabilities, assuming exchangeability.'
  4. [Throughout] There are several typos and grammatical issues: 'miss-classification' should be 'misclassification,' 'calibariation' in Section IV, 'conservativess' in Remark III.1, and 'assemsment' in the section title III-B. A careful proofreading pass is needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: InvCP is a rank-based conformal estimator whose output is not defined in terms of the target risk; the one self-citation to [22] is not load-bearing.

full rationale

The InvCP estimator in Section III-B is computed from calibration scores and the rank of s(X, I(X)) (Eq. 12), then averaged over test inputs (Eq. 11). The target quantity P(Y != Yhat(X)) is never used as an input to this computation; the estimator would be the same function of calibration scores and test scores even if the true risk were unknown. The conservativeness property (Eq. 14) is an empirical claim about the estimator, and the paper explicitly concedes in Remark III.1 that it is not guaranteed theoretically; that is a correctness and robustness limitation, not a circularity. The citation to [22] (Singh et al.) shares three authors with the present paper, but it is used only as a pointer to a regression analogue; Section III-B supplies its own derivation from CP coverage (Eqs. 6-12) and makes no load-bearing appeal to [22] for uniqueness, validity, or a forbidden alternative. The decision to use APS over LAC is described as an empirical finding on the same experiments, which is a mild selection-bias concern, but this is model selection rather than a fitted parameter renamed as a prediction: the reported delta values are out-of-sample evaluations on the independent test set Z*. No equation in the paper reduces to its own input by construction, so there is no significant circularity.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The InvCP method itself has no free parameters, which the paper highlights as an advantage. The listed free parameters belong to the comparison methods. The central claim relies on exchangeability and the empirical proxy for ground truth, both standard but unvalidated assumptions.

free parameters (2)
  • Number of bins M for HIST-BIN and ISO-REG = not specified in main results
    The main comparison figures (Figs. 1-3) do not report the bin count used for histogram binning and isotonic regression, while Fig. 5 shows strong sensitivity to M, so the comparison may be affected by an arbitrary or tuned choice.
  • Temperature T for Platt scaling = fitted on calibration set
    The temperature is optimized by minimizing negative log-likelihood on the calibration set, which is standard but is a fitted parameter in the comparison.
assumptions (3)
  • domain assumption Calibration and test sets are exchangeable i.i.d. samples from the same distribution.
    This is the standard conformal prediction assumption, stated in Section III-A: 'Under the assumption that the data is iid (could be relaxed to exchangeability)'. The method relies on this for the validity of the quantiles.
  • domain assumption The empirical misclassification rate alphaEmp on the test set is a good approximation to the true risk.
    Stated in Section IV: 'We ensure that Z* is large enough so that alphaEmp provides a good approximation to the true miss-classification rate.' The evaluation metric delta is measured against this estimate, so its accuracy bounds the conclusions.
  • domain assumption The APS score function meaningfully orders classes by model confidence.
    The paper chooses the APS score from references [27,28] and states it is paramount to performance; no theoretical justification is given for why this score is suitable for risk assessment across all models.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An In-Depth Examination of Risk Assessment in Multi-Class Classification Algorithms." pith.science (2026). https://pith.science/paper/KYR2H4ER

@misc{pith2026241204166,
  author       = {Pith},
  title        = {Pith review of: An In-Depth Examination of Risk Assessment in Multi-Class Classification Algorithms},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KYR2H4ER}},
  note         = {Machine review of arXiv:2412.04166}
}
read the original abstract

Advanced classification algorithms are being increasingly used in safety-critical applications like health-care, engineering, etc. In such applications, miss-classifications made by ML algorithms can result in substantial financial or health-related losses. To better anticipate and prepare for such losses, the algorithm user seeks an estimate for the probability that the algorithm miss-classifies a sample. We refer to this task as the risk-assessment. For a variety of models and datasets, we numerically analyze the performance of different methods in solving the risk-assessment problem. We consider two solution strategies: a) calibration techniques that calibrate the output probabilities of classification models to provide accurate probability outputs; and b) a novel approach based upon the prediction interval generation technique of conformal prediction. Our conformal prediction based approach is model and data-distribution agnostic, simple to implement, and provides reasonable results for a variety of use-cases. We compare the different methods on a broad variety of models and datasets.

Figures

Figures reproduced from arXiv: 2412.04166 by the authors.

Figure 2
Figure 2. Results for ImageNet–V15. Smaller the δ (see 13), more accurate the method. A method with δ ≥ 0 is conservative. Places365: Results for top-1 model output are shown in 3a. For AlexNet, the SMX methods provides the best results. For both the resNet18 and resNet50, HIST-BIN is the most accurate while maintaining conservativeness. For all the models, InvCP is more accurate than the PLATT method and is also conservative… view at source ↗
Figure 1
Figure 1. Results for CIFAR-100. Smaller the δ (see 13), more accurate the method. δ ≥ 0 implies conservativeness of the method. CIFAR-10 and Flowers-102: Results for both these datasets are very similar to those for CIFAR-100. One single risk￾assessment method doesn’t outperform all the other methods for all the models. Nonetheless, InvCP provides the best results in general while being free of optimization and hyper-paramet… view at source ↗
Figure 5
Figure 5. Results for ImageNet for different bin sizes [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figures from the paper (4 more)
Figure 3
Figure 3. Figure 3: Results for Places365. Smaller the δ (see 13), more accurate the method. A method with δ ≥ 0 is conservative. 5000 10000 15000 20000 25000 n 0.05 0.00 0.05 0.10 0.15 Results for ImageNet ISO-REG HIST-BIN SMX PLATT InvCP (a) Results for ImageNet 2500 5000 7500 10000 125…
Figure 4
Figure 4. Figure 4: Results for ImageNet–V1 and Places365 for different [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Results for CIFAR–10 for different values of [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Results for CIFAR–10 for different values of [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 24 canonical work pages

  1. [1]

    Image pre- processing in computer vision systems for melanoma detection

    Eugenio V ocaturo, Ester Zumpano, and Pierangelo Veltri. Image pre- processing in computer vision systems for melanoma detection. In 2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pages 2117–2124, 2018

    work page 2018

  2. [2]

    Computer aided melanoma skin cancer detection using image processing

    Shivangi Jain, Vandana jagtap, and Nitin Pise. Computer aided melanoma skin cancer detection using image processing. Procedia Computer Science, 48:735–740, 2015. International Conference on Computer, Communication and Convergence (ICCC 2015)

    work page 2015

  3. [3]

    Computer vision and digital imaging technology in melanoma detection

    Holger V oigt and Richarda Classen. Computer vision and digital imaging technology in melanoma detection. Seminars in Oncology, 29(4):308–327, 2002

    work page 2002

  4. [4]

    Scalable systems for early fault detection in wind turbines: a data driven approach

    Martin Bach-Andersen, O Winther, and B Rømer-Odgaard. Scalable systems for early fault detection in wind turbines: a data driven approach. In Annual Conference of the European Wind Energy Association , pages 17–20, 2015

    work page 2015

  5. [5]

    Deep learning for automated drivetrain fault detection

    Martin Bach-Andersen, Bo Rømer-Odgaard, and Ole Winther. Deep learning for automated drivetrain fault detection. Wind Energy, 21(1):29– 41, 2018

    work page 2018

  6. [6]

    Weinberger

    Chuan Guo, Geoff Pleiss, Yu Sun, and Kilian Q. Weinberger. On calibration of modern neural networks. In Proceedings of the 34th International Conference on Machine Learning - Volume 70 , ICML’17, page 1321–1330. JMLR.org, 2017

    work page 2017

  7. [7]

    Obtaining calibrated probability estimates from decision trees and naive bayesian classifiers

    Bianca Zadrozny and Charles Elkan. Obtaining calibrated probability estimates from decision trees and naive bayesian classifiers. In Icml, volume 1, pages 609–616, 2001

    work page 2001

  8. [8]

    Transforming classifier scores into accurate multiclass probability estimates

    Bianca Zadrozny and Charles Elkan. Transforming classifier scores into accurate multiclass probability estimates. In Proceedings of the eighth ACM SIGKDD international conference on Knowledge discovery and data mining, pages 694–699, 2002

    2002

Show all 36 references
  1. [9]

    Cooper, and Milos Hauskrecht

    Mahdi Pakdaman Naeini, Gregory F. Cooper, and Milos Hauskrecht. Obtaining well calibrated probabilities using bayesian binning. In Pro- ceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence , AAAI’15, page 2901–2907. AAAI Press, 2015

    2015

  2. [10]

    A gentle introduction to conformal prediction and distribution-free uncertainty quantification

    Anastasios N Angelopoulos and Stephen Bates. A gentle introduction to conformal prediction and distribution-free uncertainty quantification. arXiv preprint arXiv:2107.07511 , 2021

    2021 arXiv

  3. [11]

    Algorithmic learning in a random world , volume 29

    Vladimir V ovk, Alexander Gammerman, and Glenn Shafer. Algorithmic learning in a random world , volume 29. Springer, 2005

    2005

  4. [12]

    Conformalized quantile regression

    Yaniv Romano, Evan Patterson, and Emmanuel Candes. Conformalized quantile regression. Advances in neural information processing systems , 32, 2019

    2019

  5. [13]

    Image-to-image regression with distribution-free un- certainty quantification and applications in imaging

    Anastasios N Angelopoulos, Amit P Kohli, Stephen Bates, Michael I Jordan, Jitendra Malik, Thayer Alshaabi, Srigokul Upadhyayula, and Yaniv Romano. Image-to-image regression with distribution-free un- certainty quantification and applications in imaging. arXiv preprint arXiv:22...

    2022 arXiv

  6. [14]

    Angelopoulos, Jennifer Listgarten, and Michael I

    Clara Fannjiang, Stephen Bates, Anastasios N. Angelopoulos, Jennifer Listgarten, and Michael I. Jordan. Conformal prediction under feedback covariate shift for biomolecular design. Proceedings of the National Academy of Sciences , 119(43):e2204569119, 2022

    2022

  7. [15]

    Angelopoulos, Stephen Bates, Yaniv Romano, and Phillip Isola

    Swami Sankaranarayanan, Anastasios N. Angelopoulos, Stephen Bates, Yaniv Romano, and Phillip Isola. Semantic uncertainty intervals for disentangled latent spaces. arXiv preprint arXiv:2207.10074 , 2022

    2022 arXiv

  8. [16]

    Uncertainty sets for image classifiers using conformal prediction

    Anastasios Angelopoulos, Stephen Bates, Jitendra Malik, and Michael I Jordan. Uncertainty sets for image classifiers using conformal prediction. arXiv preprint arXiv:2009.14193 , 2020

    2009 arXiv

  9. [18]

    Con- formal prediction sets for ordinal classification

    Prasenjit Dey, Srujana Merugu, and Sivaramakrishnan R Kaveri. Con- formal prediction sets for ordinal classification. In A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, editors, Advances in Neural Information Processing Systems , volume 36, pages 879–899. ...

    2023

  10. [19]

    Improving expert predictions with conformal prediction

    Eleni Straitouri, Lequn Wang, Nastaran Okati, and Manuel Gomez Ro- driguez. Improving expert predictions with conformal prediction. In Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, and Jonathan Scarlett, editors, Proceedings of the 40th Inter...

    2023

  11. [20]

    Least ambiguous set-valued classifiers with bounded error levels

    Mauricio Sadinle, Jing Lei, and Larry Wasserman. Least ambiguous set-valued classifiers with bounded error levels. Journal of the American Statistical Association, 114(525):223–234, 2019

    2019

  12. [21]

    Jaws: Auditing predictive uncertainty under covariate shift

    Drew Prinster, Anqi Liu, and Suchi Saria. Jaws: Auditing predictive uncertainty under covariate shift. Advances in Neural Information Processing Systems, 35:35907–35920, 2022

    2022

  13. [22]

    Distribution-free risk assessment of regression-based machine learning algorithms, 2023

    Sukrita Singh, Neeraj Sarna, Yuanyuan Li, Yang Lin, Agni Orfanoudaki, and Michael Berger. Distribution-free risk assessment of regression-based machine learning algorithms, 2023

    2023

  14. [23]

    Classifier calibration: a survey on how to assess and improve predicted class probabilities

    Telmo Silva Filho, Hao Song, Miquel Perello-Nieto, Raul Santos- Rodriguez, Meelis Kull, and Peter Flach. Classifier calibration: a survey on how to assess and improve predicted class probabilities. Machine Learning, 112(9):3211–3260, May 2023

    2023

  15. [24]

    Measuring calibration in deep learning

    Jeremy Nixon, Michael W Dusenberry, Linchuan Zhang, Ghassen Jerfel, and Dustin Tran. Measuring calibration in deep learning. In CVPR workshops, volume 2, 2019

    2019

  16. [25]

    Inductive confidence machines for regression

    Harris Papadopoulos, Kostas Proedrou, V olodya V ovk, and Alex Gam- merman. Inductive confidence machines for regression. Lecture notes in computer science, pages 345–356, 2002

    2002

  17. [26]

    Co- variate shift adaptation by importance weighted cross validation

    Masashi Sugiyama, Matthias Krauledat, and Klaus-Robert Müller. Co- variate shift adaptation by importance weighted cross validation. Journal of Machine Learning Research , 8(5), 2007

    2007

  18. [27]

    Normalized nonconformity measures for regression conformal prediction

    Harris Papadopoulos, Alex Gammerman, and V olodya V ovk. Normalized nonconformity measures for regression conformal prediction. pages 64–69, 2008

    2008

  19. [28]

    Classification with valid and adaptive coverage

    Yaniv Romano, Matteo Sesia, and Emmanuel Candes. Classification with valid and adaptive coverage. Advances in Neural Information Processing Systems, 33:3581–3591, 2020

    2020

  20. [29]

    Cifar-100 dataset

    Alex Krizhevsky and Geoffrey Hinton. Cifar-100 dataset. https://www. cs.toronto.edu/~kriz/cifar.html, 2009

    2009

  21. [30]

    Automated flower classification over a large number of classes

    Maria-Elena Nilsback and Andrew Zisserman. Automated flower classification over a large number of classes. In 2008 Sixth Indian Conference on Computer Vision, Graphics and Image Processing , pages 722–729, 2008

    2008

  22. [31]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition , pages 248–255. IEEE, 2009

    2009

  23. [32]

    Places: A 10 million image database for scene recognition

    Bolei Zhou, Agata Lapedriza, Aditya Khosla, Aude Oliva, and Antonio Torralba. Places: A 10 million image database for scene recognition. IEEE transactions on pattern analysis and machine intelligence , 40(6):1452– 1464, 2017

    2017

  24. [33]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 770–778, 2016

    2016

  25. [34]

    Densely connected convolutional networks

    Gao Huang, Zhuang Liu, Laurens Van Der Maaten, and Kilian Q Weinberger. Densely connected convolutional networks. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 4700–4708, 2017

    2017

  26. [35]

    Imagenet classification with deep convolutional neural networks

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. Communications of the ACM , 60(6):84–90, 2017

    2017

  27. [36]

    Very deep convolutional net- works for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convolutional net- works for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014

    2014 arXiv

  28. [37]

    Pytorch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems , 32...

    2019

Pith tools

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