Pith. sign in

REVIEW 4 major objections 5 minor 57 references

Streamlining the Development of Active Learning Methods in Real-World Object Detection

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a metric computed only from labeled object crops and class counts—object-based set similarity (OSS)—predicts which active-learning methods for object detection will work before any detector is trained.

desk verdict A cheap object-crop similarity metric that correlates with mAP and stabilizes AL validation rankings, but the pre-training elimination claim rests on an unvalidated Xval-for-Dpool approximation. read the letter →

arxiv 2508.19906 v1 pith:WJDQLCB7 submitted 2025-08-27 cs.CV

Moussa Kassem Sbeyti , Nadja Klein , Michelle Karg , Christian Wirth , Sahin Albayrak This is my paper · ORCID
classification cs.CV
keywords activelearningobjectdetectionsetsimilarityuncertaintyestimationvalidationselectionautonomousdrivingdomainshiftevaluationreliability
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

Active learning for object detection normally forces a developer to train many detectors across many iterations just to find out which selection method helps; on autonomous-driving data a single training run can cost hundreds of GPU hours. This paper sets out to break that cycle by defining a pure data metric, object-based set similarity (OSS), which measures how similar one set of labeled object crops is to another using aspect ratio, texture, color, and class-count distributions. The paper argues that a training set's informativeness—how much it will raise detector mAP—and a validation set's representativeness are both governed by this similarity, so OSS can rank AL methods before training and can pick a validation subset whose rankings survive domain shift. If the claim holds, AL development becomes far cheaper and evaluation results become trustworthy enough for safety-critical use. The paper validates the metric on KITTI, BDD100K, and CODA with two detector architectures.

What carries the argument

OSS is the central object. It is defined as the inverse of a class-weighted sum, over classes, of the Jensen-Shannon divergence between two sets' feature distributions plus a smoothed class-count ratio, normalized by the 25th-percentile scale across compared sets. The feature distributions are built from object crops only: aspect ratio (shape), mean of the 2D discrete cosine transform (texture), and the mean flattened 3D color histogram (color), each modeled with Gaussian kernel density estimation so that sets of different sizes can be compared. The class-count ratio is a linear approximation of a sigmoid, and a coefficient-of-variation threshold activates class weighting only when class imb

What would settle it

Take a pool-based AL setup on a new autonomous-driving dataset, hold out a target set sampled directly from the unlabeled pool (not from the curated validation set), compute OSS rankings for several AL methods, train the detectors, and compare OSS rankings with mAP rankings; the central claim fails if Pearson correlation and Kendall's tau are no longer positive and significant, or if a randomly picked validation subset matches the alternative-domain ranking better than the OSS-selected subset.

Watch

Extended reading notes

Core claim

The paper's central claim is that the factors that make an AL training set informative are the same as the factors that make an evaluation set representative: similarity to the target domain. It introduces OSS, computed from labeled object crops per class (aspect ratios, mean 2D-DCT coefficients for texture, mean color-histogram values) and from class counts, as an estimator of that similarity. The paper reports strong linear correlations between OSS and mAP for ten AL methods across early iterations on KITTI and BDD100K, and shows that an OSS-selected subset of a validation set keeps AL-method rankings consistent across domain-shifted evaluation (Kendall's tau rising from -0.2 to 0.6 on KIT

Load-bearing premise

The load-bearing premise is that the labeled validation set Xval stands in for the unlabeled data pool Dpool: the paper approximates OSS(X_l || Dpool) with OSS(X_l || Xval), and tests OSS only against mAP measured on Xval, so if Xval is not representative of the pool, OSS can confidently rank AL methods for the wrong target.

Editorial extensions

If this is right

  • Weak AL methods can be eliminated before training: OSS takes about one minute on KITTI and 12 minutes on BDD, versus up to 806 GPU hours for a six-iteration training run, and in simulation saves up to 204 GPU hours on KITTI and 3,224 GPU hours on BDD per eliminated method.
  • Evaluation becomes more reliable under domain shift: OSS-selected validation subsets raise ranking consistency (Kendall's tau) from -0.2 to 0.6 on KITTI versus CODA and from -0.6 to 1.0 on BDD versus CODA, with no additional labeling.
  • OSS is detector-agnostic: the correlation with mAP holds for both EfficientDet and YOLOv3, so the pre-training ranking does not have to be redone when the detector changes.
  • For uncertainty-based AL, calibration and class balancing are the mechanisms that matter: they raise mAP because they increase OSS between the selected set and the target, while uncertainty separation does not help.
  • AL can reach near-full-data performance at a fraction of the data: in the case study, AL at 45% of data approaches the 100%-data mAP (70.05 on KITTI and 23.67 on BDD).

Reading between the lines

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

  • The paper's approximation step—using Xval as a proxy for the unlabeled pool—is where I would stress-test next: an independent experiment whose reference set is sampled from the actual unlabeled pool, not from Xval, would bound the error OSS can make when the validation set is unrepresentative.
  • A corollary the authors leave implicit is that OSS can double as an inexpensive domain-shift monitor: recomputing OSS between a current validation set and incoming unlabeled images would flag when the operating distribution has drifted enough to invalidate existing rankings.
  • Since OSS is computed from crops and counts only, it could also be inverted into a selection rule—choose the next images that maximize OSS to the target—unifying selection and evaluation under one objective; the paper names this as future work.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The manuscript introduces object-based set similarity (OSS), a training-free metric computed from object crops (aspect ratio, DCT, color histogram features) and class-count ratios, and uses it for two purposes: (i) ranking uncertainty-based active learning (AL) methods by predicted mAP before detector training, so that ineffective methods can be eliminated pre-training; and (ii) selecting a subset of a validation set that yields AL ranking results stable under domain shift. OSS is defined in Eq. (1) as the inverse of a class-weighted sum of per-class Jensen-Shannon divergence and a smoothed class-count ratio. The approach is validated on KITTI, BDD100K, and CODA with EfficientDet (and YOLOv3 for a cross-detector check), reporting Pearson correlations between OSS and mAP/AP and Kendall tau for ranking consistency. The paper also reports a GPU-hour cost analysis of method elimination and ablations of the OSS components.

Significance. If the central claims hold, OSS is a practically valuable tool: it requires only labeled object crops, is detector-agnostic, and could substantially reduce the computational cost of AL method development by screening methods before training and by improving evaluation-set reliability. The experimental work is extensive for the tested scope (three datasets, two detectors, ten uncertainty-based AL variants), includes a detailed component ablation, a cross-detector validation, and a transparent cost model. The main weaknesses are that the Dpool-to-Xval link is not directly validated, the class-count term dominates the OSS-mAP correlation in the ablation, and the validation-subset selection is evaluated on the same alternative set used for selection. These issues are load-bearing for the paper's broad claims, but they are fixable with additional experiments and tighter evaluation protocols rather than being fundamental flaws.

major comments (4)
  1. [III-B1 and IV-B (Fig. 3)] The pre-training claim rests on the approximation OSS(X_l||Dpool) ≈ OSS(X_l||Xval). Section III-B1 states that Dpool is mostly unlabeled and that the approximation is made; however, Fig. 3's caption and Section IV-B report correlations of OSS(X_l^{1-3}||Dpool) with mAP on Xval. This is either contradictory or the Dpool computation uses ground-truth labels that are not described. In either case, no experiment validates a pool-level target: mAP is always measured on Xval, and Xval's representativeness of Dpool is never checked. Since the advertised benefit is ranking methods for Dpool, this link must be validated directly (e.g., by computing OSS(X_l||Dpool) from full labels, or by comparing the feature/class distribution of Xval with that of Dpool).
  2. [Eq. (1)-(2) and Table III] The ablation shows that the class-count ratio r_j^(P_m) from Eq. (2) is the dominant component: removing it reduces the OSS-mAP correlation on KITTI from 0.74**/0.83*** to -0.21/0.35 (Table III), and on BDD-Xalt from 0.81***/0.89*** to 0.45/0.59. Because mAP generally increases with the number of object instances, the strong OSS-mAP correlation may largely reflect set size rather than feature similarity as a measure of informativeness. To support the central claim, the authors should control for object count—for example, by regressing mAP on r_j alone, comparing methods at matched object counts, or showing that OSS outperforms a count-only baseline. As it stands, the metric's most load-bearing term is also the one that can produce a spurious correlation.
  3. [III-B2 and IV-C] The validation-subset selection procedure chooses Xval,subset by maximizing OSS(Xval,subset||Xalt) and then reports the Kendall tau between AL rankings on Xval,subset and Xalt. Because Xalt is used both as the selection reference and as the evaluation target, the reported improvements (tau from -0.2 to 0.6 on KITTI and from -0.6 to 1.0 on BDD) are partly by construction. A fair test requires a held-out reference: select the subset using one part of Xalt (or a third evaluation set) and measure tau on the held-out part, or report cross-validated tau. Without this, the claim that OSS identifies subsets whose ranking 'generalizes under domain shift' is not supported.
  4. [IV-D and Table II] The cost-savings simulation assumes that OSS-based elimination at early iterations makes the same decisions as the post-hoc mAP-based elimination, but the paper does not specify a decision rule or report false-positive/negative rates. The text notes that low-performing methods 'typically retain their rank' (Fig. 5), yet the savings in Table II are per eliminated method rather than the expected saving under a stated threshold. Please formalize the elimination protocol and evaluate its accuracy—e.g., how often a method ranked low by OSS at i=1 would have been among the best at later iterations—so that the practical recommendation is actionable.
minor comments (5)
  1. [IV-B / Fig. 3] The caption and text should specify whether OSS(X_l||Dpool) is computed from ground-truth labels of the full pool or approximated by OSS(X_l||Xval); the current text is internally inconsistent.
  2. [III-B2] The number z of random validation subsets is manually set and no sensitivity analysis is provided. Please report how OSSz and the resulting tau depend on z.
  3. [IV-C] The description '100 subsets with 1/10 of the original size each' is ambiguous: clarify whether subsets are sampled with or without replacement and whether they are disjoint.
  4. [Sec. VI and Table I] The Discussion honestly states that the OSS-mAP correlation 'struggles with less descriptive rare classes.' This limitation should also qualify the abstract and conclusion, since Table I shows r=0.47 and r=0.10 for Traffic Light and Traffic Sign, respectively.
  5. [Notation] Small notation issues: n_det^(P_m) is used before being defined, and the symbol OSS(P_m||Q) suggests an asymmetric divergence while the text describes a similarity; a brief explanation of the notation would improve readability.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial circularity: Cw is tuned on the target tau, so the validation-subset result is in-sample; the OSS-mAP informativeness link is otherwise empirical.

  1. fitted input called prediction [Section IV-E1, 'Balancing Terms', Eq. (3) and Table V]
    "Setting Cw around 1.0 activates class weighting through wj exclusively for the strongly class-imbalanced BDD. This strengthens the correlation between similarity and τ and results in higher τ post-subsampling (see Table V). In contrast, removing wj, i.e., wj = 1 for BDD or applying it on KITTI decreases correlation."

    Cw is a free parameter in Eq. (3) that switches on the class-weighting term. The paper selects Cw=1.0 because it yields the highest τ in Table V (0.80 for BDD vs -0.66 at Cw=1.4). The reported 'representative subset' reliability is therefore the result of tuning the metric on the very outcome it is supposed to predict; it is an in-sample optimum, not an out-of-sample validation. This makes the evaluation-reliability claim partially forced by the choice of Cw, though the OSS metric still contains feature JSD and is not definitionally equal to τ.

full rationale

The core OSS-vs-mAP correlation is an empirical finding: OSS is not defined in terms of mAP, and the ablation shows a mix of feature JSD and class-count terms; no equation reduces OSS to mAP. There is no load-bearing self-citation. The main circular flavor is the manual selection of Cw on the target τ. Also, the paper's unvalidated approximation OSS(X_l||Dpool) ≈ OSS(X_l||Xval) is a generalizability gap, not circularity. Score 4 reflects partial in-sample circularity while acknowledging independent content.

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

The central claim rests on the unproven linear link between OSS and mAP, the unvalidated Xval-for-Dpool proxy, and the assumption that similarity to an alternative evaluation set confers representativeness. The only tuned free parameters are Cw and z, plus the unspecified KDE bandwidth. No new physical or conceptual entities are introduced.

free parameters (2)
  • Cw (class weighting activation threshold) = ~1.0 (activated for BDD, not for KITTI)
    Eq. (3) defines Cw as a manually defined threshold; Table V shows Cw=1.0 yields the best post-subsampling tau while Cw=1.4 degrades it, indicating the threshold was selected using the target evaluation outcome.
  • z (number of random validation subsets drawn)
    Section III-B2 says z is 'manually determined based on the size of Xval', with no stated criterion or sensitivity analysis.
assumptions (3)
  • ad hoc to paper mAP at AL iteration i is positively and linearly correlated with OSS(X_l^i || Xval).
    Section III-B1 proposes this as the theoretical link; it is not derived, only empirically demonstrated on specific datasets and methods.
  • domain assumption Since Dpool is mostly unlabeled, OSS(X_l || Dpool) can be approximated by OSS(X_l || Xval).
    Section III-B1 introduces this approximation, but the experiments never measure OSS against the true pool, so the load-bearing transfer from validation to pool is unvalidated.
  • ad hoc to paper The representativeness of Xval is measured by OSS(Xval || Xalt), so ranking reliability on Xval transfers to Xalt and to Dpool.
    Section III-B2 asserts this link and uses it to justify subsampling a validation subset that maximizes similarity to Xalt; it is not derived or tested against the full pool.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Streamlining the Development of Active Learning Methods in Real-World Object Detection." pith.science (2026). https://pith.science/paper/WJDQLCB7

@misc{pith2026250819906,
  author       = {Pith},
  title        = {Pith review of: Streamlining the Development of Active Learning Methods in Real-World Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WJDQLCB7}},
  note         = {Machine review of arXiv:2508.19906}
}
abstract

Active learning (AL) for real-world object detection faces computational and reliability challenges that limit practical deployment. Developing new AL methods requires training multiple detectors across iterations to compare against existing approaches. This creates high costs for autonomous driving datasets where the training of one detector requires up to 282 GPU hours. Additionally, AL method rankings vary substantially across validation sets, compromising reliability in safety-critical transportation systems. We introduce object-based set similarity ($\mathrm{OSS}$), a metric that addresses these challenges. $\mathrm{OSS}$ (1) quantifies AL method effectiveness without requiring detector training by measuring similarity between training sets and target domains using object-level features. This enables the elimination of ineffective AL methods before training. Furthermore, $\mathrm{OSS}$ (2) enables the selection of representative validation sets for robust evaluation. We validate our similarity-based approach on three autonomous driving datasets (KITTI, BDD100K, CODA) using uncertainty-based AL methods as a case study with two detector architectures (EfficientDet, YOLOv3). This work is the first to unify AL training and evaluation strategies in object detection based on object similarity. $\mathrm{OSS}$ is detector-agnostic, requires only labeled object crops, and integrates with existing AL pipelines. This provides a practical framework for deploying AL in real-world applications where computational efficiency and evaluation reliability are critical. Code is available at https://mos-ks.github.io/publications/.

Figures

Figures reproduced from arXiv: 2508.19906 by the authors.

Figure 1
Figure 1. The ranking of AL methods depends significantly on the composition [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. OSS quantifies both the informativeness of a training set and the representativeness of an evaluation set. By measuring the similarity between selected training sets (P) and a validation set (Q), it assesses the effectiveness of AL methods without training. Hence, it reduces computational costs by eliminating ineffective AL methods. Additionally, by measuring the similarity between validation subsets (P) and an alte… view at source ↗
Figure 3
Figure 3. Strong linear correlation visualized and quantified via [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: ). Third, as depicted in [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Bottom: Impact of various uncertainties (entropy [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 52 canonical work pages

  1. [1]

    A survey of deep active learning,

    P. Ren, Y . Xiao, X. Chang, P.-Y . Huang, Z. Li, B. B. Gupta, X. Chen, and X. Wang, “A survey of deep active learning,” ACM Computing Surveys (CSUR), vol. 54, no. 9, pp. 1–40, 2021. 1

    work page 2021

  2. [2]

    Ten years of active learning techniques and object detection: A systematic review,

    D. Garcia, J. Carias, T. Ad ˜ao, R. Jesus, A. Cunha, and L. G. Mag- alh˜aes, “Ten years of active learning techniques and object detection: A systematic review,” Applied Sciences, vol. 13, no. 19, p. 10667, 2023. 1

    work page 2023

  3. [3]

    Deep active learning for object detection,

    S. Roy, A. Unmesh, and V . P. Namboodiri, “Deep active learning for object detection,” in British Machine Vision Conference (BMVC) , 2018. 1, 2, 5, 7

    work page 2018

  4. [4]

    Advanced active learning strategies for object detection,

    S. Schmidt, Q. Rao, J. Tatsch, and A. Knoll, “Advanced active learning strategies for object detection,” in IEEE Intelligent Vehicles Symposium (IV), 2020, pp. 871–876. 1, 2

    work page 2020

  5. [5]

    Consistency-based semi-supervised active learning: Towards minimiz- ing labeling cost,

    M. Gao, Z. Zhang, G. Yu, S. ¨O. Arık, L. S. Davis, and T. Pfister, “Consistency-based semi-supervised active learning: Towards minimiz- ing labeling cost,” in European Conference on Computer Vision (ECCV), 2020, pp. 510–526. 1

    work page 2020

  6. [6]

    Active learn- ing for deep object detection via probabilistic modeling,

    J. Choi, I. Elezi, H.-J. Lee, C. Farabet, and J. M. Alvarez, “Active learn- ing for deep object detection via probabilistic modeling,” in IEEE/CVF International Conference on Computer Vision (ICCV), 2021, pp. 10 264– 10 273. 1, 2, 5, 7

    work page 2021

  7. [7]

    Localization-aware active learning for object detection,

    C.-C. Kao, T.-Y . Lee, P. Sen, and M.-Y . Liu, “Localization-aware active learning for object detection,” in Asian Conference on Computer Vision (ACCV), 2019, pp. 506–522. 1, 2

    work page 2019

  8. [8]

    Single-model uncertainties for deep learning,

    N. Tagasovska and D. Lopez-Paz, “Single-model uncertainties for deep learning,” Advances in Neural Information Processing Systems , vol. 32,

    work page

Show all 57 references
  1. [9]

    Uncertainty estimates as data selection criteria to boost omni- supervised learning,

    L. Venturini, A. T. Papageorghiou, J. A. Noble, and A. I. Nambu- rete, “Uncertainty estimates as data selection criteria to boost omni- supervised learning,” in Medical Image Computing and Computer As- sisted Intervention (MICCAI) , 2020, pp. 689–698. 1

    2020

  2. [10]

    Aleatoric and epistemic uncertainty in machine learning: An introduction to concepts and methods,

    E. H ¨ullermeier and W. Waegeman, “Aleatoric and epistemic uncertainty in machine learning: An introduction to concepts and methods,”Machine Learning, vol. 110, pp. 457–506, 2021. 1

    2021

  3. [11]

    Complementing semi-supervised learning with uncertainty quantification,

    E. Kazemi, “Complementing semi-supervised learning with uncertainty quantification,” arXiv preprint arXiv:2207.12131 , 2022. 1

    2022 arXiv

  4. [12]

    How to measure un- certainty in uncertainty sampling for active learning,

    V .-L. Nguyen, M. H. Shaker, and E. H ¨ullermeier, “How to measure un- certainty in uncertainty sampling for active learning,”Machine Learning, vol. 111, no. 1, pp. 89–122, 2022. 1

    2022

  5. [13]

    A deeper look into aleatoric and epistemic uncertainty disentanglement,

    M. Valdenegro-Toro and D. S. Mori, “A deeper look into aleatoric and epistemic uncertainty disentanglement,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) , 2022, pp. 1508–1516. 1

    2022

  6. [14]

    Deep ensemble bayesian active learning: Address- ing the mode collapse issue in monte carlo dropout via ensembles,

    R. Pop and P. Fulop, “Deep ensemble bayesian active learning: Address- ing the mode collapse issue in monte carlo dropout via ensembles,”arXiv preprint arXiv:1811.03897, 2018. 1, 2

    2018 arXiv

  7. [15]

    Towards dynamic and scalable active learning with neural architecture adaption for object detection,

    F. Tang, D. Wei, C. Jiang, H. Xu, A. Zhang, W. Zhang, H. Lu, and C. Xu, “Towards dynamic and scalable active learning with neural architecture adaption for object detection,” in British Machine Vision Conference (BMVC), 2021. 1, 2

    2021

  8. [16]

    Active learning on a budget: Opposite strategies suit high and low budgets,

    G. Hacohen, A. Dekel, and D. Weinshall, “Active learning on a budget: Opposite strategies suit high and low budgets,” arXiv preprint arXiv:2202.02794, 2022. 1, 2 10

    2022 arXiv

  9. [17]

    Active learning for deep object detection,

    C.-A. Brust, C. K ¨ading, and J. Denzler, “Active learning for deep object detection,” arXiv preprint arXiv:1809.09875 , 2018. 1, 2, 5

    2018 arXiv

  10. [18]

    Talisman: targeted active learning for object detection with rare classes and slices using submodular mutual information,

    S. Kothawade, S. Ghosh, S. Shekhar, Y . Xiang, and R. Iyer, “Talisman: targeted active learning for object detection with rare classes and slices using submodular mutual information,” in European Conference on Computer Vision (ECCV) , 2022, pp. 1–16. 1, 2, 3

    2022

  11. [19]

    Entropy-based active learning for object detection with progressive diversity constraint,

    J. Wu, J. Chen, and D. Huang, “Entropy-based active learning for object detection with progressive diversity constraint,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 9397–9406. 1, 2

    2022

  12. [20]

    Scalable active learning for object detection,

    E. Haussmann, M. Fenzi, K. Chitta, J. Ivanecky, H. Xu, D. Roy, A. Mit- tel, N. Koumchatzky, C. Farabet, and J. M. Alvarez, “Scalable active learning for object detection,” in IEEE Intelligent Vehicles Symposium (IV), 2020, pp. 1430–1435. 1, 2, 3, 5

    2020

  13. [21]

    The pascal visual object classes (voc) challenge,

    M. Everingham, L. Van Gool, C. K. Williams, J. Winn, and A. Zisser- man, “The pascal visual object classes (voc) challenge,” International Journal of Computer Vision , vol. 88, pp. 303–338, 2010. 1

    2010

  14. [22]

    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 European Conference on Computer Vision (ECCV) , 2014, pp. 740–755. 1, 3, 5

    2014

  15. [23]

    Parting with illusions about deep active learning,

    S. Mittal, M. Tatarchenko, ¨O. C ¸ ic ¸ek, and T. Brox, “Parting with illusions about deep active learning,” arXiv preprint arXiv:1912.05361 , 2019. 1, 3

    1912 arXiv

  16. [24]

    Albench: a framework for evaluating active learning in object detection,

    Z. Feng, S. Zhang, R. Takezoe, W. Hu, M. Chandraker, L.-J. Li, V . K. Narayanan, and X. Wang, “Albench: a framework for evaluating active learning in object detection,” arXiv preprint arXiv:2207.13339, 2022. 1, 2

    2022 arXiv

  17. [25]

    Coda: A real-world road corner case dataset for object detection in autonomous driving,

    K. Li, K. Chen, H. Wang, L. Hong, C. Ye, J. Han, Y . Chen, W. Zhang, C. Xu, D.-Y . Yeunget al., “Coda: A real-world road corner case dataset for object detection in autonomous driving,” in European Conference on Computer Vision (ECCV) , 2022, pp. 406–423. 2, 5, 6

    2022

  18. [26]

    Practical obstacles to deploying active learning,

    D. Lowell, Z. C. Lipton, and B. C. Wallace, “Practical obstacles to deploying active learning,” arXiv preprint arXiv:1807.04801 , 2018. 2

    2018 arXiv

  19. [27]

    Towards rapid prototyping and comparability in active learning for deep object detection,

    T. Riedlinger, M. Schubert, K. Kahl, H. Gottschalk, and M. Rottmann, “Towards rapid prototyping and comparability in active learning for deep object detection,” arXiv preprint arXiv:2212.10836 , 2022. 2, 7

    2022 arXiv

  20. [28]

    A survey on active learning: State-of-the- art, practical challenges and research directions,

    A. Tharwat and W. Schenck, “A survey on active learning: State-of-the- art, practical challenges and research directions,” Mathematics, vol. 11, no. 4, p. 820, 2023. 2

    2023

  21. [29]

    Learning loss for active learning,

    D. Yoo and I. S. Kweon, “Learning loss for active learning,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 93–102. 2

    2019

  22. [30]

    A re-balancing strategy for class-imbalanced classification based on instance difficulty,

    S. Yu, J. Guo, R. Zhang, Y . Fan, Z. Wang, and X. Cheng, “A re-balancing strategy for class-imbalanced classification based on instance difficulty,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 70–79. 2

    2022

  23. [31]

    Smote: synthetic minority over-sampling technique,

    N. V . Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “Smote: synthetic minority over-sampling technique,” Journal of Artificial Intel- ligence Research (JAIR), vol. 16, pp. 321–357, 2002. 2

    2002

  24. [32]

    Consistency-based active learning for object detection,

    W. Yu, S. Zhu, T. Yang, and C. Chen, “Consistency-based active learning for object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2022, pp. 3951–

    2022

  25. [33]

    Multiple instance active learning for object detection,

    T. Yuan, F. Wan, M. Fu, J. Liu, S. Xu, X. Ji, and Q. Ye, “Multiple instance active learning for object detection,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2021, pp. 5330–

    2021

  26. [34]

    Discrete cosine transform,

    N. Ahmed, T. Natarajan, and K. R. Rao, “Discrete cosine transform,” IEEE transactions on Computers , vol. 100, no. 1, pp. 90–93, 2006. 4

    2006

  27. [35]

    Derivations for linear algebra and optimization,

    J. Duchi, “Derivations for linear algebra and optimization,” Berkeley, California, vol. 3, no. 1, pp. 2325–5870, 2007. 4

    2007

  28. [36]

    D. W. Scott, Multivariate density estimation: theory, practice, and visualization. John Wiley & Sons, 2015. 4

    2015

  29. [37]

    SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python,

    P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson, K. J. Millman, N. Mayorov, A. R. J. Nelson, E. Jones, R. Kern, E. Larson, C. J. Carey,˙I. Polat, Y . Feng...

    2020

  30. [38]

    Kullback-leibler divergence estimation of continuous distributions,

    F. P ´erez-Cruz, “Kullback-leibler divergence estimation of continuous distributions,” in IEEE International Symposium on Information Theory , 2008, pp. 1666–1670. 4

    2008

  31. [39]

    C. M. Bishop, Neural networks for pattern recognition . Oxford university press, 1995. 5

    1995

  32. [40]

    Efficientdet: Scalable and efficient object detection,

    M. Tan, R. Pang, and Q. V . Le, “Efficientdet: Scalable and efficient object detection,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 10 778–10 787. 5

    2020

  33. [41]

    GitHub - google/automl: Google Brain AutoML,

    Google, “GitHub - google/automl: Google Brain AutoML,” https:// github.com/google/automl, 2020, accessed: 2024-06-10. 5

    2020

  34. [42]

    Are we ready for autonomous driving? the kitti vision benchmark suite,

    A. Geiger, P. Lenz, and R. Urtasun, “Are we ready for autonomous driving? the kitti vision benchmark suite,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2012, pp. 3354–

    2012

  35. [43]

    Bdd100k: A diverse driving dataset for heterogeneous multitask learning,

    F. Yu, H. Chen, X. Wang, W. Xian, Y . Chen, F. Liu, V . Madhavan, and T. Darrell, “Bdd100k: A diverse driving dataset for heterogeneous multitask learning,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 2636–2645. 5

    2020

  36. [44]

    Settles, Active learning literature survey

    B. Settles, Active learning literature survey . University of Wisconsin- Madison Department of Computer Sciences, 2009. 5

    2009

  37. [45]

    Efficient object localization using convolutional networks,

    J. Tompson, R. Goroshin, A. Jain, Y . LeCun, and C. Bregler, “Efficient object localization using convolutional networks,” in IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR) , 2015, pp. 648–656. 5

    2015

  38. [46]

    Uncertainty estimation in deep neural object detectors for autonomous driving,

    M. Stoycheva, “Uncertainty estimation in deep neural object detectors for autonomous driving,” Master’s thesis, KTH, School of Electrical Engineering and Computer Science (EECS), 2021. 5

    2021

  39. [47]

    What uncertainties do we need in bayesian deep learning for computer vision?

    A. Kendall and Y . Gal, “What uncertainties do we need in bayesian deep learning for computer vision?” Advances in Neural Information Processing Systems, vol. 30, 2017. 5

    2017

  40. [48]

    Overcoming the limitations of localization uncertainty: Efficient and exact non-linear post-processing and calibration,

    M. Kassem Sbeyti, M. Karg, C. Wirth, A. Nowzad, and S. Albayrak, “Overcoming the limitations of localization uncertainty: Efficient and exact non-linear post-processing and calibration,” in Machine Learn- ing and Knowledge Discovery in Databases: Research Track. ECML- PKDD., 2...

    2023

  41. [49]

    Vii. note on regression and inheritance in the case of two parents,

    K. Pearson, “Vii. note on regression and inheritance in the case of two parents,” Proceedings of the royal society of London , vol. 58, no. 347- 352, pp. 240–242, 1895. 5

  42. [50]

    A new measure of rank correlation,

    M. G. Kendall, “A new measure of rank correlation,” Biometrika, vol. 30, no. 1-2, pp. 81–93, 1938. 5

    1938

  43. [51]

    On initial pools for deep active learning,

    A. L. Chandra, S. V . Desai, C. Devaguptapu, and V . N. Balasubramanian, “On initial pools for deep active learning,” in Advances in Neural Information Processing Systems Workshop , 2021, pp. 14–32. 7

    2021

  44. [52]

    Localization- based active learning (local) for object detection in 3d point clouds,

    A. Moses, S. Jakkampudi, C. Danner, and D. Biega, “Localization- based active learning (local) for object detection in 3d point clouds,” in Geospatial Informatics XII , vol. 12099, 2022, pp. 44–58. 7

    2022

  45. [53]

    Gpu pricing,

    Google Cloud, “Gpu pricing,” https://cloud.google.com/compute/ gpus-pricing, 2025, accessed: 2025-07-06. 8

    2025

  46. [54]

    Very deep convolutional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv preprint arXiv:1409.1556 , 2015. 8

    2015 arXiv

  47. [55]

    Vgg16 and vgg19,

    Keras, “Vgg16 and vgg19,” https://keras.io/api/applications/vgg, 2024, accessed: 2024-06-10. 8

    2024

  48. [56]

    Yolov3: An incremental improvement,

    A. Farhadi and J. Redmon, “Yolov3: An incremental improvement,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), vol. 1804, 2018, pp. 1–6. 9

    2018

  49. [57]

    GitHub - yolov3 in tensorflow 2.0,

    Z. Zhang, “GitHub - yolov3 in tensorflow 2.0,” https://github.com/ zzh8829/yolov3-tf2, 2019, accessed: 2025-01-06. 9

    2019

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.