Skip to main content
SpringerLink
Log in

WMBAL: weighted minimum bounds for active learning

  • Published:
Applied Intelligence Aims and scope Submit manuscript

Abstract

In the present study, aimed at reliably acquiring difficult samples for object detection models from massive raw data, we propose a novel difficult sample mining strategy based on active learning with Weighted Minimum Bounds (WMB). To accurately gauge the difficulty of samples for object detection models, we introduce the concept of weighted minimum bounds. Uniquely, the metric for measuring sample difficulty includes the classification discrepancy within detection frames and a weight factor derived from the Average Precision (AP) of the object detection model on the val set. Additionally, we introduce the Don’t Care Area (DCA) to capture the uncertainty in localization tasks for object detection models. The DCA is utilized only during the data mining and training phases, ensuring that no additional time is incurred during inference. Furthermore, we propose a periodic and phased framework based on active learning for mining difficult samples, which can progressively identify challenging samples from unlabeled data and perform iterative optimization. To evaluate the effectiveness of our methods, we have collected the VANJEE-Image dataset and the VANJEE-PointCloud dataset from real-world scenarios. We empirically demonstrate the superiority of our approach, which outperforms traditional active learning methods on both image detection and point cloud detection datasets. The code and datasets are available at https://github.com/sharkls/WMBAL-Weighted-Minimum-Bounds-for-Active-Learning.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Algorithm 1
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8

Similar content being viewed by others

References

  1. Parvaneh A, Abbasnejad E, Teney D, Haffari GR, Van Den Hengel A, Shi JQ (2022) Active learning by feature mixing, pp 12237–12246

  2. Hemmer P, Kühl N, Schöffer J (2022) Deal: deep evidential active learning for image classification. Deep Learn Appl 3:171–192

    Google Scholar 

  3. Yang C, An Z, Cai L, Xu Y (2022) Mutual contrastive learning for visual representation learning, vol 36, no 3, pp 3045–3053

  4. Köksal A, Schick T, Schütze H (2022) Meal: stable and active learning for few-shot prompting. arXiv:2211.08358

  5. Zheng Y, Gao Y, Lu S, Mosalam KM (2022) Multistage semisupervised active learning framework for crack identification, segmentation, and measurement of bridges. Comput Aided Civ Infrastruct Eng 37(9):1089–1108

    Article  Google Scholar 

  6. Lu Q, Wei L (2021) Multiscale superpixel-based active learning for hyperspectral image classification. IEEE Geosci Remote Sens Lett 19:1–5

    Google Scholar 

  7. Cao X, Yao J, Xu Z, Meng D (2020) Hyperspectral image classification with convolutional neural network and active learning. IEEE Trans Geosci Remote Sens 58(7):4604–4616

    Article  ADS  Google Scholar 

  8. Kothawade S, Savarkar A, Iyer V, Ramakrishnan G, Iyer R (2022) Clinical: targeted active learning for imbalanced medical image classification, 119–129. Springer

  9. Sener O, Savarese S (2017) Active learning for convolutional neural networks: a core-set approach. arXiv:1708.00489

  10. Agarwal S, Arora H, Anand S, Arora C (2020) Contextual diversity for active learning, 137–153. Springer

  11. Xie B, Yuan L, Li S, Liu CH, Cheng X (2022) Towards fewer annotations: active learning via region impurity and prediction uncertainty for domain adaptive semantic segmentation, 8068–8078

  12. Yadav CS, Pradhan MK, Gangadharan SMP, Chaudhary JK, Singh J, Khan AA, Haq MA, Alhussen A, Wechtaisong C, Imran H et al (2022) Multi-class pixel certainty active learning model for classification of land cover classes using hyperspectral imagery. Electronics 11(17):2799

    Article  Google Scholar 

  13. Huang H, Liu Z, Chen CP, Zhang Y (2023) Hyperspectral image classification via active learning and broad learning system. Appl Intell 53(12):15683–15694

    Article  Google Scholar 

  14. Zheng Y, Gao Y, Lu S, Mosalam KM (2022) Multistage semisupervised active learning framework for crack identification, segmentation, and measurement of bridges. Comput Aided Civ Infrastruct Eng 37(9):1089–1108

    Article  Google Scholar 

  15. Wang S, Li Y, Ma K, Ma R, Guan H, Zheng Y (2020) Dual adversarial network for deep active learning, 680–696. Springer

  16. Ho DJ, Agaram NP, Schüffler PJ, Vanderbilt CM, Jean M-H, Hameed MR, Fuchs TJ (2020) Deep interactive learning: an efficient labeling approach for deep learning-based osteosarcoma treatment response assessment, 540–549. Springer

  17. Kim D-J, Cho JW, Choi J, Jung Y, Kweon IS (2021) Single-modal entropy based active learning for visual question answering. arXiv:2110.10906

  18. Ash JT, Zhang C, Krishnamurthy A, Langford J, Agarwal A (2019) Deep batch active learning by diverse, uncertain gradient lower bounds. arXiv:1906.03671

  19. Yi JSK, Seo M, Park J, Choi D-G (2022) Pt4al: using self-supervised pretext tasks for active learning, 596–612. Springer

  20. Xu Y, Ma L, Xiao W (2019) Active learning with spatial distribution based semi-supervised extreme learning machine for multiclass classification, 1–5. IEEE

  21. Goodfellow I, Pouget-Abadie J, Mirza M, Xu B, Warde-Farley D, Ozair S, Courville A, Bengio Y (2014) Generative adversarial nets. Advances in neural information processing systems 27

  22. Cho JW, Kim D-J, Jung Y, Kweon IS (2022) Mcdal: maximum classifier discrepancy for active learning. IEEE Trans Neural Netw Learn Syst 1–11

  23. Yoo D, Kweon IS (2019) Learning loss for active learning, 93–102

  24. Zhang B, Li L, Yang S, Wang S, Zha Z-J, Huang Q (2020) State-relabeling adversarial active learning, 8756–8765

  25. Aghdam HH, Gonzalez-Garcia A, Weijer Jvd, López AM (2019) Active learning for deep detection neural networks, 3672–3680

  26. Angluin D (1988) Queries and concept learning. Mach Learn 2:319–342

    Article  MathSciNet  Google Scholar 

  27. Dagan I, Engelson SP (1995) Committee-based sampling for training probabilistic classifiers, 150–157

  28. King RD, Whelan KE, Jones FM, Reiser PG, Bryant CH, Muggleton SH, Kell DB, Oliver SG (2004) Functional genomic hypothesis generation and experimentation by a robot scientist. Nature 427(6971):247–252

    Article  ADS  CAS  PubMed  Google Scholar 

  29. Krishnamurthy V (2002) Algorithms for optimal scheduling and management of hidden markov model sensors. IEEE Trans Signal Process 50(6):1382–1397

    Article  ADS  Google Scholar 

  30. Schröder C, Niekler A, Potthast M (2021) Revisiting uncertainty-based query strategies for active learning with transformers. arXiv:2107.05687

  31. Buchert F, Navab N, Kim ST (2022) Exploiting diversity of unlabeled data for label-efficient semi-supervised active learning, 2063–2069. IEEE

  32. Shui C, Zhou F, Gagné C, Wang B (2020) Deep active learning: unified and principled method for query and training, 1308–1318. PMLR

  33. Dor LE, Halfon A, Gera A, Shnarch E, Dankin L, Choshen L, Danilevsky M, Aharonov R, Katz Y, Slonim N (2020) Active learning for bert: an empirical study, 7949–7962

  34. Kumar P, Gupta A (2020) Active learning query strategies for classification, regression, and clustering: a survey. J Comput Sci Technol 35:913–945

    Article  Google Scholar 

  35. Yu W, Zhu S, Yang T, Chen C (2022) Consistency-based active learning for object detection, 3951–3960

  36. Tong S (2001) Active learning: theory and applications

  37. Xie J, Ma Z, Lei J, Zhang G, Xue J-H, Tan Z-H, Guo J (2021) Advanced dropout: a model-free methodology for Bayesian dropout optimization. IEEE Trans Pattern Anal Mach Intell 44(9):4605–4625

    Google Scholar 

  38. Ho DJ, Agaram NP, Schüffler PJ, Vanderbilt CM, Jean M-H, Hameed MR, Fuchs TJ (2020) Deep interactive learning: an efficient labeling approach for deep learning-based osteosarcoma treatment response assessment, 540–549. Springer

  39. Schumann R, Rehbein I (2019) Active learning via membership query synthesis for semi-supervised sentence classification, 472–481

  40. Vaith A, Taetz B, Bleser G (2020) Uncertainty based active learning with deep neural networks for inertial gait analysis, 1–8. IEEE

  41. Sinha S, Ebrahimi S, Darrell T (2019) Variational adversarial active learning, 5972–5981

  42. Wang X, Xiang X, Zhang B, Liu X, Zheng J, Hu Q (2022) Weakly supervised object detection based on active learning. Neural Process Lett 54(6):5169–5183

  43. Wang S, Li Y, Ma K, Ma R, Guan H, Zheng Y (2020) Dual adversarial network for deep active learning, 680–696. Springer

  44. Zhang B, Li L, Yang S, Wang S, Zha Z-J, Huang Q (2020) State-relabeling adversarial active learning, 8756–8765

  45. Gal Y, Ghahramani Z (2016) Dropout as a Bayesian approximation: representing model uncertainty in deep learning, 1050–1059. PMLR

Download references

Author information

Authors and Affiliations

  1. The Department of Intelligent Internet Connection, Beijing VANJEE Technology Co., Ltd, Beijing, 100085, China

    Shuai Lu, Jiaxi Zheng, Zhi Li & Xuerui Dai

Authors
  1. Shuai Lu
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Jiaxi Zheng
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Zhi Li
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Xuerui Dai
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Xuerui Dai.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Lu, S., Zheng, J., Li, Z. et al. WMBAL: weighted minimum bounds for active learning. Appl Intell 54, 2551–2563 (2024). https://doi.org/10.1007/s10489-024-05328-x

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10489-024-05328-x

Keywords

Search

Navigation