pith. sign in

arxiv: 2604.20083 · v1 · submitted 2026-04-22 · 💻 cs.LG · cs.CV

Energy-Based Open-Set Active Learning for Object Classification

Zongyao Lyu , William J. Beksi This is my paper

Pith reviewed 2026-05-10 01:41 UTC · model grok-4.3

classification 💻 cs.LG cs.CV
keywords active learningopen-set recognitionenergy-based modelsobject classificationannotation efficiencydeep learning
0
0 comments X

The pith

A dual-stage energy framework filters unknown classes then scores known samples for labeling in open-set active learning.

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

The paper develops a method for active learning when the unlabeled pool contains both known and unknown classes. It uses one energy-based model to separate the two by assigning lower energy to known samples and higher energy to unknowns, then a second energy-based model to rank the remaining known samples by how useful they are to label. This avoids wasting the annotation budget on irrelevant classes. On image and 3D object benchmarks the approach yields higher final accuracy and better use of each labeling round than prior active-learning techniques that assume a closed set.

Core claim

The authors propose a novel dual-stage energy-based framework for open-set active learning consisting of an energy-based known/unknown separator and an energy-based sample scorer. By using the energy landscape to assign lower energy to known-class samples and higher energy to unknown-class samples, the framework filters out unknowns from the unlabeled pool before selecting informative known samples, resulting in superior performance on CIFAR-10, CIFAR-100, TinyImageNet, and ModelNet40 benchmarks.

What carries the argument

Dual-stage energy-based framework using two EBMs: a known/unknown separator that removes unknown-class samples and a sample scorer that ranks the informativeness of the retained known samples.

If this is right

  • Annotation budgets are spent only on samples from the target classes.
  • Each labeling iteration improves classification accuracy more than closed-set methods in open environments.
  • The same energy scores can be reused across multiple active-learning rounds without retraining from scratch.
  • The approach scales to both 2D image and 3D point-cloud classification tasks.

Where Pith is reading between the lines

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

  • The separation step could be paired with any existing informativeness scorer rather than only an energy-based one.
  • If the energy landscape generalizes across domains, the framework might apply to sequential data or text classification with emerging classes.
  • A natural test would be to measure how performance changes when the fraction of unknown classes in the pool is varied systematically.

Load-bearing premise

Energy-based models can reliably assign lower energy to known-class samples and higher energy to unknown-class samples in the unlabeled pool without any prior exposure to the unknown classes.

What would settle it

An experiment in which the energy separator produces similar energy values for known and unknown samples, so that the method selects roughly the same proportion of unknowns as standard active learning and shows no gain in annotation efficiency or final accuracy.

Figures

Figures reproduced from arXiv: 2604.20083 by William J. Beksi, Zongyao Lyu.

Figure 1
Figure 1. Figure 1: An illustration of the OSAL problem. The unlabeled set contains samples from [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: An overview of the EB-OSAL framework. The proposed approach follows a [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Object classification accuracy on CIFAR-10 with mismatch ratios of 20%, 30%, [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Object classification accuracy on CIFAR-100 with mismatch ratios of 20%, 30%, [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Object classification accuracy on TinyImageNet with mismatch ratios of 20%, [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Object classification overall accuracy on ModelNet40 with mismatch ratios of [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: An ablation study on CIFAR-100 with a mismatch ratio of 20%. [PITH_FULL_IMAGE:figures/full_fig_p016_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: An ablation study on ModelNet40 with a mismatch ration of 20%. [PITH_FULL_IMAGE:figures/full_fig_p017_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: The impact of margin parameters on CIFAR100. [PITH_FULL_IMAGE:figures/full_fig_p017_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: The impact of margin parameters on ModelNet40. [PITH_FULL_IMAGE:figures/full_fig_p018_10.png] view at source ↗
read the original abstract

Active learning (AL) has emerged as a crucial methodology for minimizing labeling costs in deep learning by selecting the most valuable samples from a pool of unlabeled data for annotation. Traditional AL operates under a closed-set assumption, where all classes in the dataset are known and consistent. However, real-world scenarios often present open-set conditions in which unlabeled data contains both known and unknown classes. In such environments, standard AL techniques struggle. They can mistakenly query samples from unknown categories, leading to inefficient use of annotation budgets. In this paper, we propose a novel dual-stage energy-based framework for open-set AL. Our method employs two specialized energy-based models (EBMs). The first, an energy-based known/unknown separator, filters out samples likely to belong to unknown classes. The second, an energy-based sample scorer, assesses the informativeness of the filtered known samples. Using the energy landscape, our models distinguish between data points from known and unknown classes in the unlabeled pool by assigning lower energy to known samples and higher energy to unknown samples, ensuring that only samples from classes of interest are selected for labeling. By integrating these components, our approach ensures efficient and targeted sample selection, maximizing learning impact in each iteration. Experiments on 2D (CIFAR-10, CIFAR-100, TinyImageNet) and 3D (ModelNet40) object classification benchmarks demonstrates that our framework outperforms existing approaches, achieving superior annotation efficiency and classification performance in open-set environments.

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

2 major / 2 minor

Summary. The paper proposes a novel dual-stage energy-based framework for open-set active learning in object classification. It employs two energy-based models (EBMs): an energy-based known/unknown separator that filters samples from unknown classes in the unlabeled pool by assigning them higher energy, and an energy-based sample scorer that evaluates the informativeness of the remaining known-class samples. The approach is evaluated on 2D benchmarks (CIFAR-10, CIFAR-100, TinyImageNet) and a 3D benchmark (ModelNet40), with the claim that it achieves superior annotation efficiency and classification performance compared to existing methods under open-set conditions.

Significance. If the central claims hold, the work would represent a meaningful contribution to active learning by addressing the practical challenge of unknown classes in unlabeled pools, which standard closed-set AL methods mishandle. The dual use of EBMs for both separation and scoring is a coherent integration of energy-based modeling into the AL loop, and the inclusion of 3D data (ModelNet40) alongside 2D benchmarks strengthens the evaluation scope. The paper correctly identifies the inefficiency of querying unknowns and proposes a targeted filtering mechanism.

major comments (2)
  1. [Method (dual-stage energy-based framework)] The core mechanism of the energy-based known/unknown separator (described in the method overview) relies on the unverified assumption that an EBM trained solely on known-class data will assign reliably lower energy to in-distribution samples and higher energy to any unknown-class samples in the pool. No training objective, loss formulation, architectural details, or regularization (e.g., contrastive terms or gradient penalties) are specified to guarantee this separation. This is load-bearing for the open-set claim, as standard EBM energy functions on CIFAR-100 and TinyImageNet frequently fail to produce clean OOD thresholds without explicit mechanisms, directly undermining the filtering step before scoring.
  2. [Experiments] The experimental claims (Abstract and Experiments section) state that the framework outperforms existing approaches on CIFAR-10, CIFAR-100, TinyImageNet, and ModelNet40 with superior annotation efficiency and classification performance, yet no quantitative results, specific metrics (e.g., accuracy curves, annotation savings percentages), baselines, or ablation studies are reported. Without these, the superiority assertion cannot be verified and the effect sizes remain unknown.
minor comments (2)
  1. [Abstract] Grammatical error in the Abstract: 'Experiments on 2D ... and 3D ... benchmarks demonstrates' should read 'demonstrate' to match the plural subject.
  2. [Abstract] The repeated phrasing in the Abstract about assigning 'lower energy to known samples and higher energy to unknown samples' could be condensed to improve readability without loss of meaning.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We address each major comment point by point below and commit to a thorough revision that strengthens the technical presentation and empirical support.

read point-by-point responses

  1. Referee: [Method (dual-stage energy-based framework)] The core mechanism of the energy-based known/unknown separator (described in the method overview) relies on the unverified assumption that an EBM trained solely on known-class data will assign reliably lower energy to in-distribution samples and higher energy to any unknown-class samples in the pool. No training objective, loss formulation, architectural details, or regularization (e.g., contrastive terms or gradient penalties) are specified to guarantee this separation. This is load-bearing for the open-set claim, as standard EBM energy functions on CIFAR-100 and TinyImageNet frequently fail to produce clean OOD thresholds without explicit mechanisms, directly undermining the filtering step before scoring.

    Authors: We agree that the current method overview provides only a high-level description of the energy-based known/unknown separator and does not specify the training objective, loss formulation, architectural details, or regularization terms needed to ensure reliable separation. This detail is essential for substantiating the open-set filtering step. In the revised manuscript we will add a dedicated subsection that defines the EBM energy function, the precise training loss (including any contrastive or regularization components such as gradient penalties), network architecture choices, and hyper-parameters. We will also include supporting empirical analysis (e.g., energy distribution plots on held-out known and unknown samples) to demonstrate that the separation holds on the CIFAR-100 and TinyImageNet benchmarks. revision: yes

  2. Referee: [Experiments] The experimental claims (Abstract and Experiments section) state that the framework outperforms existing approaches on CIFAR-10, CIFAR-100, TinyImageNet, and ModelNet40 with superior annotation efficiency and classification performance, yet no quantitative results, specific metrics (e.g., accuracy curves, annotation savings percentages), baselines, or ablation studies are reported. Without these, the superiority assertion cannot be verified and the effect sizes remain unknown.

    Authors: We acknowledge that the manuscript draft does not contain the quantitative experimental results, performance curves, annotation-efficiency percentages, baseline comparisons, or ablation studies referenced in the abstract and summary. This omission prevents independent verification of the claimed improvements. In the revised version we will expand the Experiments section to include all requested quantitative results: accuracy-vs-annotation-budget curves, explicit annotation savings percentages, comparisons against the listed baselines, and ablation studies on each of the four benchmarks (CIFAR-10, CIFAR-100, TinyImageNet, ModelNet40). revision: yes

Circularity Check

0 steps flagged

No circularity: empirical method with standard EBM assumption tested on benchmarks

full rationale

The paper proposes a dual-stage EBM framework where one model separates known/unknown samples and the other scores informativeness, with the separation property (lower energy for known, higher for unknown) presented as following from training exclusively on known-class data. This is a standard EBM modeling choice, not a self-definitional loop or fitted input renamed as prediction. No equations, uniqueness theorems, or self-citations are shown to reduce the central claim to its own inputs by construction. The work is self-contained as an empirical proposal validated on CIFAR-10/100, TinyImageNet, and ModelNet40 without load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

The framework rests on the domain assumption that EBMs can be trained to separate known and unknown distributions via energy values, plus two newly introduced model components.

axioms (1)
  • domain assumption Energy-based models assign lower energy to samples from the known class distribution and higher energy to unknown class samples
    Invoked to justify the known/unknown separator and sample scorer.
invented entities (2)
  • Energy-based known/unknown separator no independent evidence
    purpose: Filters samples likely from unknown classes in the unlabeled pool
    New component introduced by the paper; no independent evidence provided.
  • Energy-based sample scorer no independent evidence
    purpose: Assesses informativeness of the remaining known samples
    Second new component introduced by the paper; no independent evidence provided.

pith-pipeline@v0.9.0 · 5562 in / 1219 out tokens · 45614 ms · 2026-05-10T01:41:28.734811+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

54 extracted references · 54 canonical work pages

  1. [1]

    In: Proceed- ings of the Advances in Neural Information Processing Systems

    Alliegro, A., Cappio Borlino, F., Tommasi, T.: 3dos: Towards 3d open set learning- benchmarking and understanding semantic novelty detection on point clouds. In: Proceed- ings of the Advances in Neural Information Processing Systems. vol. 35, pp. 21228–21240 (2022)

    work page 2022

  2. [2]

    In: Proceedings of the International Confer- ence on Learning Representations (2020)

    Ash, J.T., Zhang, C., Krishnamurthy, A., Langford, J., Agarwal, A.: Deep batch active learn- ing by diverse, uncertain gradient lower bounds. In: Proceedings of the International Confer- ence on Learning Representations (2020)

    work page 2020

  3. [3]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Beluch, W.H., Genewein, T., Nürnberger, A., Köhler, J.M.: The power of ensembles for active learning in image classification. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 9368–9377 (2018)

    work page 2018

  4. [4]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Bendale, A., Boult, T.E.: Towards open set deep networks. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 1563–1572 (2016)

    work page 2016

  5. [5]

    In: Proceedings of the International Conference on 3D Vision

    Cen, J., Yun, P., Cai, J., Wang, M.Y ., Liu, M.: Open-set 3d object detection. In: Proceedings of the International Conference on 3D Vision. pp. 869–878. IEEE (2021)

    work page 2021

  6. [6]

    In: Proceedings of the Advances in Neural Information Processing Systems

    Chen, S., Huang, L.K., Schwarz, J.R., Du, Y ., Wei, Y .: Secure out-of-distribution task gener- alization with energy-based models. In: Proceedings of the Advances in Neural Information Processing Systems. vol. 36 (2023)

    work page 2023

  7. [7]

    Journal of Artificial Intelligence Research4, 129–145 (1996)

    Cohn, D.A., Ghahramani, Z., Jordan, M.I.: Active learning with statistical models. Journal of Artificial Intelligence Research4, 129–145 (1996)

    work page 1996

  8. [8]

    In: Proceedings of the British Machine Vision Confer- ence

    Cui, R., Qiu, S., Anwar, S., Zhang, J., Barnes, N.: Energy-based residual latent transport for unsupervised point cloud completion. In: Proceedings of the British Machine Vision Confer- ence. British Machine Vision Association and Society for Pattern Recognition (2022)

    work page 2022

  9. [9]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Deng, J., Dong, W., Socher, R., Li, L.J., Li, K., Fei-Fei, L.: Imagenet: A large-scale hierar- chical image database. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 248–255 (2009)

    work page 2009

  10. [10]

    In: Pro- ceedings of the Advances in Neural Information Processing Systems

    Du, Y ., Mordatch, I.: Implicit generation and modeling with energy based models. In: Pro- ceedings of the Advances in Neural Information Processing Systems. vol. 32 (2019)

    work page 2019

  11. [11]

    https://github.com/robotic-vision-lab/Energy-Based-Open-Set-Active-Learning-For- Object-Classification

    work page

  12. [12]

    In: Pro- ceedings of the International Conference on Machine Learning

    Gal, Y ., Islam, R., Ghahramani, Z.: Deep bayesian active learning with image data. In: Pro- ceedings of the International Conference on Machine Learning. pp. 1183–1192 (2017)

    work page 2017

  13. [13]

    In: Proceedings of the British Machine Vision Conference

    Ge, Z., Demyanov, S., Chen, Z., Garnavi, R.: Generative openmax for multi-class open set classification. In: Proceedings of the British Machine Vision Conference. British Machine Vision Association and Society for Pattern Recognition (2017)

    work page 2017

  14. [14]

    In: Proceedings of the International Conference on Learning Representations (2020)

    Grathwohl, W., Wang, K.C., Jacobsen, J.H., Duvenaud, D., Norouzi, M., Swersky, K.: Your classifier is secretly an energy based model and you should treat it like one. In: Proceedings of the International Conference on Learning Representations (2020)

    work page 2020

  15. [15]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Gustafsson, F.K., Danelljan, M., Schon, T.B.: Accurate 3d object detection using energy- based models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 2855–2864 (2021)

    work page 2021

  16. [16]

    In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 770–778 (2016)

    work page 2016

  17. [17]

    IEEE Transactions on Pattern Analysis and Machine Intelligence10(36), 1936– 1949 (2014)

    Huang, S.J., Jin, R., Zhou, Z.H.: Active learning by querying informative and representative examples. IEEE Transactions on Pattern Analysis and Machine Intelligence10(36), 1936– 1949 (2014)

    work page 1936

  18. [18]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Joshi, A.J., Porikli, F., Papanikolopoulos, N.: Multi-class active learning for image clas- sification. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 2372–2379 (2009) 14 Zongyao Lyu and William J. Beksi

    work page 2009

  19. [19]

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

    Kim, Y ., Yim, J., Yun, J., Kim, J.: Nlnl: Negative learning for noisy labels. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 101–110 (2019)

    work page 2019

  20. [20]

    In: Pro- ceedings of the IEEE/CVF International Conference on Computer Vision

    Kong, S., Ramanan, D.: Opengan: Open-set recognition via open data generation. In: Pro- ceedings of the IEEE/CVF International Conference on Computer Vision. pp. 813–822 (2021)

    work page 2021

  21. [21]

    Krizhevsky, A.: Learning multiple layers of features from tiny images. Tech. rep., University of Toronto, Toronto, Ontario (2009)

    work page 2009

  22. [22]

    CS 231N7(7), 3 (2015)

    Le, Y ., Yang, X.: Tiny imagenet visual recognition challenge. CS 231N7(7), 3 (2015)

    work page 2015

  23. [23]

    Predicting Structured Data1, 0 (2006)

    LeCun, Y ., Chopra, S., Hadsell, R., Ranzato, M., Huang, F.J.: A tutorial on energy-based learning. Predicting Structured Data1, 0 (2006)

    work page 2006

  24. [24]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li, J., Dong, Q.: Open-set semantic segmentation for point clouds via adversarial prototype framework. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 9425–9434 (2023)

    work page 2023

  25. [25]

    In: Proceed- ings of the Advances in Neural Information Processing Systems

    Liu, W., Wang, X., Owens, J., Li, Y .: Energy-based out-of-distribution detection. In: Proceed- ings of the Advances in Neural Information Processing Systems. vol. 33, pp. 21464–21475 (2020)

    work page 2020

  26. [26]

    In: Proceedings of the International Conference on Pattern Recognition

    Lyu, Z., Beksi, W.J.: Semi-supervised variational adversarial active learning via learning to rank and agreement-based pseudo labeling. In: Proceedings of the International Conference on Pattern Recognition. pp. 1–16. Springer (2024)

    work page 2024

  27. [27]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision Workshops

    Lyu, Z., Gutierrez, N.B., Beksi, W.J.: Metamax: Improved open-set deep neural networks via weibull calibration. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision Workshops. pp. 439–443 (2023)

    work page 2023

  28. [28]

    In: Proceedings of the International Conference on Machine Learning

    Nguyen, H.T., Smeulders, A.: Active learning using pre-clustering. In: Proceedings of the International Conference on Machine Learning. p. 79 (2004)

    work page 2004

  29. [29]

    In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Ning, K.P., Zhao, X., Li, Y ., Huang, S.J.: Active learning for open-set annotation. In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 41–49 (2022)

    work page 2022

  30. [30]

    In: Proceedings of the Advances in Neural Information Processing Systems

    Park, D., Shin, Y ., Bang, J., Lee, Y ., Song, H., Lee, J.G.: Meta-query-net: Resolving purity- informativeness dilemma in open-set active learning. In: Proceedings of the Advances in Neural Information Processing Systems. vol. 35, pp. 31416–31429 (2022)

    work page 2022

  31. [31]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Qi, C.R., Su, H., Mo, K., Guibas, L.J.: Pointnet: Deep learning on point sets for 3d classifi- cation and segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 652–660 (2017)

    work page 2017

  32. [32]

    In: Pro- ceedings of the AAAI Conference on Artificial Intelligence (2024)

    Safaei, B., VS, V ., de Melo, C.M., Patel, V .M.: Entropic open-set active learning. In: Pro- ceedings of the AAAI Conference on Artificial Intelligence (2024)

    work page 2024

  33. [33]

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

    Samet, N., Siméoni, O., Puy, G., Ponimatkin, G., Marlet, R., Lepetit, V .: You never get a second chance to make a good first impression: Seeding active learning for 3d semantic seg- mentation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 18445–18457 (2023)

    work page 2023

  34. [34]

    Synthesis Lec- tures on Computer Vision7(1), 1–131 (2017)

    Scheirer, W.J.: Extreme value theory-based methods for visual recognition. Synthesis Lec- tures on Computer Vision7(1), 1–131 (2017)

    work page 2017

  35. [35]

    IEEE Transactions on Pattern Analysis and Machine Intelligence36(11), 2317–2324 (2014)

    Scheirer, W.J., Jain, L.P., Boult, T.E.: Probability models for open set recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence36(11), 2317–2324 (2014)

    work page 2014

  36. [36]

    IEEE Transactions on Pattern Analysis and Machine Intelligence35(7), 1757–1772 (2012)

    Scheirer, W.J., de Rezende Rocha, A., Sapkota, A., Boult, T.E.: Toward open set recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence35(7), 1757–1772 (2012)

    work page 2012

  37. [37]

    In: Proceedings of the International Conference on Learning Representations (2018)

    Sener, O., Savarese, S.: Active learning for convolutional neural networks: A core-set ap- proach. In: Proceedings of the International Conference on Learning Representations (2018)

    work page 2018

  38. [38]

    Settles, B.: Active learning literature survey. Tech. rep., University of Wisconsin-Madison Department of Computer Science (2009) Energy-Based Open-Set Active Learning for Object Classification 15

    work page 2009

  39. [39]

    In: Proceedings of the ACM International Conference on Multimedia

    Shao, F., Luo, Y ., Liu, P., Chen, J., Yang, Y ., Lu, Y ., Xiao, J.: Active learning for point cloud semantic segmentation via spatial-structural diversity reasoning. In: Proceedings of the ACM International Conference on Multimedia. pp. 2575–2585 (2022)

    work page 2022

  40. [40]

    In: International Conference on Medical Image Computing and Computer-Assisted Intervention

    Top, A., Hamarneh, G., Abugharbieh, R.: Active learning for interactive 3d image segmen- tation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 603–610 (2011)

    work page 2011

  41. [41]

    In: Proceedings of the International Joint Conference on Neural Networks

    Wang, D., Shang, Y .: A new active labeling method for deep learning. In: Proceedings of the International Joint Conference on Neural Networks. pp. 112–119 (2014)

    work page 2014

  42. [42]

    In: Proceedings of the International Conference on Machine Learning

    Welling, M., Teh, Y .W.: Bayesian learning via stochastic gradient langevin dynamics. In: Proceedings of the International Conference on Machine Learning. pp. 681–688 (2011)

    work page 2011

  43. [43]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wu, T.H., Liu, Y .C., Huang, Y .K., Lee, H.Y ., Su, H.T., Huang, P.C., Hsu, W.H.: Redal: Region-based and diversity-aware active learning for point cloud semantic segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 15510–15519 (2021)

    work page 2021

  44. [44]

    In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition

    Wu, Z., Song, S., Khosla, A., Yu, F., Zhang, L., Tang, X., Xiao, J.: 3d shapenets: A deep representation for volumetric shapes. In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition. pp. 1912–1920 (2015)

    work page 1912

  45. [45]

    In: Proceedings of the Advances in Neural Information Processing Systems

    Xie, B., Li, S., Guo, Q., Liu, C., Cheng, X.: Annotator: A generic active learning baseline for lidar semantic segmentation. In: Proceedings of the Advances in Neural Information Processing Systems. vol. 36, pp. 48444–48458 (2023)

    work page 2023

  46. [46]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Xie, B., Yuan, L., Li, S., Liu, C.H., Cheng, X., Wang, G.: Active learning for domain adap- tation: An energy-based approach. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 36, pp. 8708–8716 (2022)

    work page 2022

  47. [47]

    In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Xie, J., Xu, Y ., Zheng, Z., Zhu, S.C., Wu, Y .N.: Generative pointnet: Deep energy-based learning on unordered point sets for 3d generation, reconstruction and classification. In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 14976–14985 (2021)

    work page 2021

  48. [48]

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

    Xu, Z., Yuan, B., Zhao, S., Zhang, Q., Gao, X.: Hierarchical point-based active learning for semi-supervised point cloud semantic segmentation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 18098–18108 (2023)

    work page 2023

  49. [49]

    arXiv preprint arXiv:2311.02601 (2023)

    Yamauchi, R., Sakurai, J., Furukawa, R., Matsubayashi, T.: Optimizing implicit neural rep- resentations from point clouds via energy-based models. arXiv preprint arXiv:2311.02601 (2023)

    work page arXiv 2023

  50. [50]

    Applied Intelligence52(11), 13250–13267 (2022)

    Yan, X., Nazmi, S., Gebru, B., Anwar, M., Homaifar, A., Sarkar, M., Gupta, K.D.: A clustering-based active learning method to query informative and representative samples. Applied Intelligence52(11), 13250–13267 (2022)

    work page 2022

  51. [51]

    International Journal of Computer Vision113, 113–127 (2015)

    Yang, Y ., Ma, Z., Nie, F., Chang, X., Hauptmann, A.G.: Multi-class active learning by uncer- tainty sampling with diversity maximization. International Journal of Computer Vision113, 113–127 (2015)

    work page 2015

  52. [52]

    In: Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition

    Yoshihashi, R., Shao, W., Kawakami, R., You, S., Iida, M., Naemura, T.: Classification- reconstruction learning for open-set recognition. In: Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition. pp. 4016–4025 (2019)

    work page 2019

  53. [53]

    In: Proceedings of the International Conference on Machine Learning

    Zhai, S., Cheng, Y ., Lu, W., Zhang, Z.: Deep structured energy based models for anomaly detection. In: Proceedings of the International Conference on Machine Learning. pp. 1100– 1109 (2016)

    work page 2016

  54. [54]

    In: Proceedings of the European Conference on Com- puter Vision

    Zong, C.C., Wang, Y .W., Ning, K.P., Ye, H.B., Huang, S.J.: Bidirectional uncertainty-based active learning for open-set annotation. In: Proceedings of the European Conference on Com- puter Vision. pp. 127–143. Springer (2024) 16 Zongyao Lyu and William J. Beksi Appendix This appendix provides additional experimental results and analyses to complement the...

    work page 2024