The DOI matches its Crossref record; 47 of the 49 references it cites were found in a registry or an index.
Active Learning for Deep Object Detection via Probabilistic Modeling
Open accessAbstract
Active learning aims to reduce labeling costs by selecting only the most informative samples on a dataset. Few existing works have addressed active learning for object detection. Most of these methods are based on multiple models or are straightforward extensions of classification methods, hence estimate an image’s informativeness using only the classification head. In this paper, we propose a novel deep active learning approach for object detection. Our approach relies on mixture density networks that estimate a probabilistic distribution for each localization and classification head’s output. We explicitly estimate the aleatoric and epistemic uncertainty in a single forward pass of a single model. Our method uses a scoring function that aggregates these two types of uncertainties for both heads to obtain every image’s informativeness score. We demonstrate the efficacy of our approach in PASCAL VOC and MS-COCO datasets. Our approach outperforms single-model based methods and performs on par with multi-model based methods at a fraction of the computing cost. Code is available at https://github.com/NVlabs/AL-MDN.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source doesn’t support the claim as written.
In the citing paper
“In the last one, where we use 7 , 000 samples, 5 , 000 of which are actively labeled, our method outperforms the random baseline by 9 . 1% and the best existing active learning methods by more than 2 . 8% [5].”
From Not All Labels Are Equal: Rationalizing The Labeling Costs for Training Object Detection (2021) · cites this paper as [5]
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“For MS-COCO, in Fig. 4a-b, we observe that in the first active learning cycle, our method outperforms the random baseline by 5 . 8% , the best-performing AL method by 2 . 7% [5], the semi-supervised method by 5 . 4% , and the ensembles by 5% .”
From Not All Labels Are Equal: Rationalizing The Labeling Costs for Training Object Detection (2021) · cites this paper as [5]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“We observe that this difference is maintained in the other cycles, including the last active learning cycle where our method outperforms the semisupervised method, multi-model methods [9, 12] and the best AL method [5] by 1 . 6% .”
From Not All Labels Are Equal: Rationalizing The Labeling Costs for Training Object Detection (2021) · cites this paper as [5]
The cited source says what the claim attributes to it.
In the citing paper
“In the second cycle, our approach outperforms all the other methods, including PM [5], by almost 4% or more.”
From Not All Labels Are Equal: Rationalizing The Labeling Costs for Training Object Detection (2021) · cites this paper as [5]
The cited source says what the claim attributes to it.
In the citing paper
“Compared to state-of-the-art methods [4-8], our approach yields up to a 7 . 7% and 7% relative mAP improvement for PASCAL-VOC and MS-COCO, respectively.”
From Not All Labels Are Equal: Rationalizing The Labeling Costs for Training Object Detection (2021) · cites this paper as [5]
The cited source says what the claim attributes to it.
In the citing paper
“The random sampling strategy is mostly used for comparison purposes, grounded on to the principle that AL-based methods have the potential to perform better (e.g., [5,48,52,60]).”
From Ten Years of Active Learning Techniques and Object Detection: A Systematic Review (2023) · cites this paper as [48]
The cited source says what the claim attributes to it.
In the citing paper
“The classifiers used for uncertainty estimation ignore the location information [48].”
From Ten Years of Active Learning Techniques and Object Detection: A Systematic Review (2023) · cites this paper as [48]
The cited source says what the claim attributes to it.
In the citing paper
“White methods, based on multi-models (like classifier assemblies which increase the number of parameters [48]), have a high computing cost, while black methods may not need many parameters.”
From Ten Years of Active Learning Techniques and Object Detection: A Systematic Review (2023) · cites this paper as [48]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 1×
“We compare the performance of SSD using different backbones such as Resnet-34 and Resnet-50 [17], and Faster-RCNN [32] detector trained using our actively sampled dataset.”
MethodVery Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 1×
“Experimental settings: We employ Single Shot MultiBox Detector (SSD) [31], which is widely used in active learning studies [33, 40], with a VGG-16 backbone [37].”
MethodFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2016.2577031 · Cited by 73,472
A Mathematical Theory of Communication
Shannon, C. E. (1948) · The MIT Press eBooks · doi:10.1002/j.1538-7305.1948.tb00917.x · Cited by 50,387
Microsoft COCO: Common Objects in Context
Lin, T.-Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Dollár, P., & Zitnick, C. L. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10602-1_48 · Cited by 45,460
(2016) · Cited by 31,590
Feature Pyramid Networks for Object Detection
Lin, T.-Y., Dollar, P., Girshick, R., He, K., Hariharan, B., & Belongie, S. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.106 · Cited by 30,939 · cited 3×
“21 Faster-RCNN [32] Resnet-50-FPN [29] 72.”
MethodInfluentialGirshick, R. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.169 · Cited by 28,105
Auto-Encoding Variational Bayes
Kingma, D. P., & Welling, M. (2013) · arXiv · doi:10.48550/arxiv.1312.6114 · Cited by 20,067 · cited 1×
“2, and obtain the class probability distribution for the k-th mixture by using the reparameterization trick [25] of applying Gaussian noise and variance Σp to μ k p [24]:”
MethodThe pascal visual object classes (VOC) challenge
Everingham, M., Van Gool, L., Williams, C. K. I., Winn, J., & Zisserman, A. (2010) · International Journal of Computer Vision · doi:10.1007/s11263-009-0275-4 · Cited by 19,599 · cited 2×
“We demonstrate the benefits of our approach on PASCAL VOC [9] and MS-COCO [30] in a single-stage architecture such as SSD [31], and show generalized performance in a two-stage architecture such as Faster-RCNN [32].”
Method
Research works printed but not linked: 4
- 10.5555/3295222.3295309 · DOI
- 10.48550/arXiv.1405.0312 · DOI · Paper page
- 10.1007/978-3-319-46448-0_2 · DOI · Paper page
- 10.1145/584091.584093 · DOI
Cited by
Sorted by citations
Tiny object detection with context enhancement and feature purification
Xiao, J., Guo, H., Zhou, J. Y., Zhao, T., Yu, Q., Chen, Y., & Wang, Z. (2022) · Expert Systems with Applications · doi:10.1016/j.eswa.2022.118665 · Cited by 153
A comprehensive survey on deep active learning in medical image analysis
Wang, H., Jin, Q., Li, S., Liu, S., Wang, M., & Song, Z. (2024) · Medical Image Analysis · doi:10.1016/j.media.2024.103201 · Cited by 157
Wang, K., Shen, C., Li, X., & Lǚ, J. (2025) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2025.3532803 · Cited by 77
“There are inherently many unknown classes in the unknown world, but it can be reduced by active learning [62], [63] and other measures.”
Semantic Segmentation with Active Semi-Supervised Learning
Rangnekar, A., Kanan, C., & Hoffman, M. J. (2022) · IEEE Workshop/Winter Conference on Applications of Computer Vision · doi:10.1109/wacv56688.2023.00591 · Cited by 44
“Query-computing methods are further classified into applying Bayesian inference [37,28,45,24,48,76,32], ensemble learning [6,68,33,63,18], diversity learning [72,85,21,77,22,53,11] and representation learning [55,41,1,42,12,19,29,30,84,87,75,62,23].”
BackgroundMethodUnbiased Scene Graph Generation in Videos
Nag, S., Min, K., Tripathi, S., & Roy–Chowdhury, A. K. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.02184 · Cited by 45
“The class-specific aleatoric and epistemic uncertainty, for the sample z i are computed as follows: Therefore, by using a GMM head, we are modeling the inherent uncertainty associated with the data from a Bayesian perspective [8, 9, 28].”
BackgroundMethodPlug and Play Active Learning for Object Detection
Yang, C., Huang, L., & Crowley, E. J. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.01684 · Cited by 46
“We follow previous work [10, 42, 46] and use a batch active learning setting, i.e. at each round we query a batch of images for oracle annotation instead of a single sample.”
BackgroundMethodInfluentialWang, N., Chang, H., & Zhang, D. (2023) · Water Resources Research · doi:10.1029/2022wr033644 · Cited by 34
“…has been applied in various fields, such as inverse materials and devices design (Chen & Gu, 2020; Pestourie et al., 2020; Tang, Kojima, et al., 2020), speech emotion recognition (Abdelwahab & Busso, 2019; Zhao & Ma, 2013), object detection (Aghdam et al., 2019; Choi et al., 2021), and so on.”
MethodLyu, M., Zhou, J., Chen, H., Huang, Y., Yu, D., Li, Y., Guo, Y., Guo, Y., Xiang, L., & Ding, G. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.02276 · Cited by 29
ALWOD: Active Learning for Weakly-Supervised Object Detection
Wang, Y., Ilić, V., Li, J., Kisačanin, B., & Pavlović, V. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00594 · Cited by 28
“We mainly focus on comparing against the state-of-the-art FSOD baselines: Faster-RCNN [36], SSD [25], and Sparse DETR [38]; WSOD baselines: WSDDN [2], OICR [46], C-MIDN [11], WSOD2 [57], MIST(+Reg) [37], CASD [19], W2N [18], and D2DF2WOD [52]; SSOD baselines: BCNet [30], OAM [1], Unbiased teacher v2 [27], and Label Ma…”
BackgroundMethodInfluentialDeep Active Learning for Computer Vision: Past and Future
Takezoe, R., Liu, X., Mao, S., Chen, M. T., Feng, Z., Zhang, S., & Wang, X. (2023) · APSIPA Transactions on Signal and Information Processing · doi:10.1561/116.00000057 · Cited by 26
“[84] learn a Gaussian Mixture Model (GMM) for the network outputs including both localization and classification.”
BackgroundTowards Building Self-Aware Object Detectors via Reliable Uncertainty Quantification and Calibration
Öksüz, K., Joy, T., & Dokania, P. K. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00894 · Cited by 27
Active Learning Strategies for Weakly-Supervised Object Detection
Vo, H. V., Siméoni, O., Gidaris, S., Bursuc, A., Pérez, P., & Ponce, J. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20056-4_13 · Cited by 25
Region-Based Hybrid Collaborative Perception for Connected Autonomous Vehicles
Liu, P., Wang, Z., Yu, G., Zhou, B., & Chen, P. (2023) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2023.3324439 · Cited by 24
“Previousstudies[44],[45]conductedthattheoutputofthemodel canalsobeexpressedastheprobabilitydistributiontocapture theuncertaintyforthenoisycondition.”
A Lightweight YOLOv5-MNE Algorithm for SAR Ship Detection
Pang, L., Li, B., Zhang, F., Meng, X., & Zhang, L. (2022) · Sensors · doi:10.3390/s22187088 · Cited by 22
Active Learning for Single-Stage Object Detection in UAV Images
Yamani, A. Z., Alyami, A., Luqman, H., Ghanem, B., & Giancola, S. (2024) · IEEE Workshop/Winter Conference on Applications of Computer Vision · doi:10.1109/wacv57701.2024.00187 · Cited by 20
“Choi et al .”
Bae, Y., Nam, K., & Kang, S. (2024) · Computers in Industry · doi:10.1016/j.compind.2024.104122 · Cited by 21
Improving the Intra-class Long-Tail in 3D Detection via Rare Example Mining
Jiang, C. M., Najibi, M., Qi, C. R., Zhou, Y., & Anguelov, D. (2022) · Lecture Notes in Computer Science · doi:10.1007/978-3-031-20080-9_10 · Cited by 21
The Road to Safety: A Review of Uncertainty and Applications to Autonomous Driving Perception
Araújo, B., Teixeira, J. F., Fonseca, J., Cerqueira, R., & Beco, S. C. (2024) · Entropy · doi:10.3390/e26080634 · Cited by 28
Ten Years of Active Learning Techniques and Object Detection: A Systematic Review
Garcia, D., Carias, J., Adão, T., Jesus, R., Cunha, A., & Magalhães, L. G. (2023) · Applied Sciences · doi:10.3390/app131910667 · Cited by 17
Zhang, X., Feng, Y., Zhang, S., Wang, N., Mei, S., & He, M. (2023) · Remote Sensing · doi:10.3390/rs15112928 · Cited by 16
Monocular 3D Object Detection with LiDAR Guided Semi Supervised Active Learning
Hekimoglu, A., Schmidt, M., & Marcos-Ramiro, A. (2023) · IEEE Workshop/Winter Conference on Applications of Computer Vision · doi:10.1109/wacv57701.2024.00234 · Cited by 22
“[9] estimate the aleatoric and epistemic uncertainty for both classification and localization and combine the uncertainties in a single selection score.”
BackgroundMethodInfluentialKecor: Kernel Coding Rate Maximization for Active 3D Object Detection
Luo, Y., Chen, Z., Fang, Z., Zhang, Z., Baktashmotlagh, M., & Huang, Z. H. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01676 · Cited by 15
“Research on active learning for object detection [9, 23, 30, 52, 69, 71] has not bee as widespread as that for image classification, due in part to the challenges in quantifying aleatoric and epistemic uncertainties in bounding box regression.”
BackgroundMethodInfluentialStream-Based Active Distillation for Scalable Model Deployment
Manjah, D., Cacciarelli, D., Benkedadra, M., Standaert, B., de Hertaing, G. R., Macq, B. M. M., Galland, S., & Vleeschouwer, C. D. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00528 · Cited by 14
“Query strategies based on localization tightness and stability [21], mixture density networks [12], and a unified box regression and classification metric [39] have also been proposed.”
MethodActive Pointly-Supervised Instance Segmentation
Tang, C., Xie, L., Zhang, G., Zhang, X., Tian, Q. C., & Hu, X. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19815-1_35 · Cited by 13
Optical Flow Training Under Limited Label Budget via Active Learning
Yuan, S., Sun, X., Kim, H. H., Yu, S., & Tomasi, C. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20047-2_24 · Cited by 12
Heterogeneous Diversity Driven Active Learning for Multi-Object Tracking
Li, R., Zhang, B., Liu, J., Liu, W., Zhao, J. C., & Teng, Z. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00911 · Cited by 12
“[1, 8, 39, 20], object detection [45, 44, 10], semantic segmentation [7, 28, 33], and so on.”
BackgroundBayesian Active Learning for Sim-to-Real Robotic Perception
Feng, J., Lee, J., Durner, M., & Triebel, R. (2021) · IEEE/RJS International Conference on Intelligent RObots and Systems · doi:10.1109/iros47612.2022.9982175 · Cited by 16
“Meanwhile, uncertainty based approaches [19], [25], [31] are also able to achieve competitive performances in the field of object detection.”
BackgroundPlausible Uncertainties for Human Pose Regression
Bramlage, L., Karg, M. E., & Curio, C. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01389 · Cited by 12
Active Domain Adaptation with False Negative Prediction for Object Detection
Nakamura, Y., Ishii, Y., & Yamashita, T. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.02719 · Cited by 15
“Active Learning (AL) for Object Detection [5, 10, 22, 24, 26, 30, 42, 45, 46, 53, 56, 58] methods aim to select effective samples for training within a limited labeling budget.”
BackgroundMethodEfficient Active Learning Strategies for Monocular 3D Object Detection
Hekimoglu, A., Schmidt, M., Marcos-Ramiro, A., & Rigoll, G. (2022) · IEEE Intelligent Vehicles Symposium · doi:10.1109/iv51971.2022.9827454 · Cited by 12
One Step Closer to Unbiased Aleatoric Uncertainty Estimation
Wang, Z., Martin, Z., Das, S., Weng, T.-W., Megretski, A., Daniel, L., & Nguyen, L. M. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i15.29627 · Cited by 11
Nani, P., Das, S., & Dey, S. (2025) · Evolutionary Intelligence · doi:10.1007/s12065-025-01061-7 · Cited by 11
“However, probabilistic models [28] indicate these discrepancies are generally minor, offering useful hints for identification.”
Im, Y.-J., Chang, T., Park, S., & Kim, H. (2025) · Waste Management · doi:10.1016/j.wasman.2025.01.028 · Cited by 9
Active Object Detection with Epistemic Uncertainty and Hierarchical Information Aggregation
Park, Y.-H., Kim, S., Choi, W., Han, D.-J., & Moon, J. J. (2022) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw56347.2022.00306 · Cited by 9
Development of robust detector using the weather deep generative model for outdoor monitoring system
Jin, K., Kang, K.-S., Shin, B.-K., Kwon, J., Jang, S., Kim, Y., & Ryu, H.-G. (2023) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.120984 · Cited by 12
Introspective Perception for Long-Term Aerial Telemanipulation With Virtual Reality
Lee, J., Balachandran, R., Kondak, K., Coelho, A., Stefano, M. D., Humt, M., Feng, J., Asfour, T., & Triebel, R. (2024) · IEEE Transactions on Field Robotics · doi:10.55417/fr.2023010 · Cited by 8
“Defining the combination function comb ( · ) as a weighted sum or max operation (Choi et al., 2021), and also the aggregation function agg ( · ) as a sum or average operation (Roy et al., 2018), it is chosen;:”
UTrack: Multi-object Tracking with Uncertain Detections
Solano‐Carrillo, E., Sattler, F., Alex, A., Klein, A., Costa, B. P., Rodríguez, Á. B., & Stoppe, J. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-91585-7_14 · Cited by 11
Pixel Is All You Need: Adversarial Trajectory-Ensemble Active Learning for Salient Object Detection
Wu, Z., Wang, L., Wang, W., Xia, Q., Chen, C., Hao, A., & Li, S. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v37i3.25390 · Cited by 8
Probabilistic Model-Based Active Learning with Attention Mechanism for Fish Species Recognition
Nabi, M. M., Shah, C., Alaba, S. Y., Prior, J. H., Campbell, M. D., Wallace, F., Moorhead, R. J., II, & Ball, J. E. (2023) · MTS/IEEE OCEANS Conference (OCEANS) · doi:10.23919/oceans52994.2023.10337403 · Cited by 8
“The model depends on the Gaussian Mixture Model (GMM) to estimate the probability distribution [12], [13].”
MethodInfluentialMulti-Task Consistency for Active Learning
Hekimoglu, A., Friedrich, P., Zimmer, W., Schmidt, M., Marcos-Ramiro, A., & Knoll, A. C. (2023) · 2023 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW) · doi:10.1109/iccvw60793.2023.00366 · Cited by 17
“AL methods for object detection measure uncertainty of a detected box through either classification or localization uncertainty [8], [16], [21]–[24].”
MethodVLMine: Long-Tail Data Mining with Vision Language Models
Ye, M., Meyer, G. P., Zhang, Z., Park, D., Mustikovela, S. K., Chai, Y., & Wolff, E. M. (2025) · IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw65960.2025.00120 · Cited by 8
“For that reason, the long-tail signal provided by our approach tends to be complementary to the signals obtained from traditional model-based mining algorithms (4; 8; 27).”
Semi-Automated Framework for Digitalizing Multi-Product Warehouses with Large Scale Camera Arrays
Higashiura, K., Yokoyama, K., Asai, Y., Shimosato, H., Kano, K., Katayama, S., Urano, K., Yonezawa, T., & Kawaguchi, N. (2024) · IEEE International Conference on Pervasive Computing and Communications · doi:10.1109/percom59722.2024.10494498 · Cited by 7
Employing feature mixture for active learning of object detection
Zhang, L., Lam, S.-K., Luo, D., & Wu, X. H. (2024) · Neurocomputing · doi:10.1016/j.neucom.2024.127883 · Cited by 6
Wu, Z.-Y., Wang, W., Wang, L., Li, Y., Lv, F.-M., Xia, Q., Chen, C., Hao, A.-M., & Li, S. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3476683 · Cited by 7
Ada-iD: Active Domain Adaptation for Intrusion Detection
Han, F., Ye, P., Duan, S., & Wang, L. (2024) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3664647.3681272 · Cited by 8
“Currently, ADA achieves encouraging performance in multiple multimedia tasks, e.g. , image classification [36], semantic segmentation [23, 29, 37, 40], object detection [1, 5, 15, 25, 28, 41], image captioning [43], refractive error detection [8] and autonomous driving [12].”
Active Supervised Cross-Modal Retrieval
Zhang, H., Yang, Y., Qi, F., Qian, S., & Xu, C. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3550526 · Cited by 13
“The predictive model is obtained after the iteration stops. classification [17], [18], [19], [20], [21], object detection [22], [23],[24],semanticsegmentation[25],[26],[27],[28],andother non-visualareas[29],[30],[31],[32],[33],[34],[35],[36].”
Assessing domain gap for continual domain adaptation in object detection
Doan, A.-D., Nguyen, B. L., Gupta, S., Reid, I. D., Wagner, M., & Chin, T.-J. (2023) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2023.103885 · Cited by 6
MD3D: Mixture-Density-Based 3D Object Detection in Point Clouds
Choi, J., Song, Y., Kim, Y., Yoo, J., & Kwak, N. (2022) · IEEE Access · doi:10.1109/access.2022.3210108 · Cited by 5
“The MDN was also utilized to model the multi-modal nature of object detection and human pose estimation [35], and to address active learning for object detection [36].”
MethodSelf-citationWang, Y., Yu, Z., Liu, S., Zhou, Z., & Guo, B. (2023) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3580815 · Cited by 6
“Here we utilize an estimation function [10] to predict the annotation difficulty of H-CNN.”
MethodSemi-Path: An interactive semi-supervised learning framework for gigapixel pathology image analysis
Lai, Z., Chauhan, J., Chen, D., Dugger, B. N., Cheung, S.-C. S., & Chuah, C. (2024) · Smart Health · doi:10.1016/j.smhl.2024.100474 · Cited by 5
ALStereo: Active learning for stereo matching
Zhang, J., Li, J., Gu, M., Yu, X., Zheng, J. G., Bai, X. P., & Hancock, E. R. (2025) · Pattern Recognition · doi:10.1016/j.patcog.2025.111480 · Cited by 4
Efficient Remote Sensing in Agriculture via Active Learning and Opt-HRDNet
Chen, D., Xiao, S., Xi, M., Dai, L., & Yang, J. (2024) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2024.3369189 · Cited by 4
“There is no doubt that remote sensing detection relying on large-scale datasets has made great progress in recent years [25], [26].”
BackgroundPrediction Accuracy & Reliability: Classification and Object Localization Under Distribution Shift
Diet, F., Sbeyti, M. K., & Karg, M. E. (2024) · Studies in big data · doi:10.1007/978-3-031-66842-5_9 · Cited by 4
“Within the field of perception for autonomous driving, uncertainty quantification has been applied to 1) classification tasks such as semantic segmentation [1], 2) regression tasks such as depth [7, 8] and optical flow estimation [6], and 3) combined tasks such as object detection [2, 3, 9, 10].”
BackgroundMethodUncertainty-driven active developmental learning
Hu, Q., Ji, L., Wang, Y., Zhao, S., & Lin, Z. (2024) · Pattern Recognition · doi:10.1016/j.patcog.2024.110384 · Cited by 4
Hierarchical Active Learning for Low-Altitude Drone-View Object Detection
Hu, H., Han, T., Wang, Y., Zhong, W., Yue, J., & Zan, P. (2024) · International Journal of Computer Vision · doi:10.1007/s11263-024-02228-y · Cited by 4
“Choi et al. (2021) employs mixture density networks to estimate both aleatoric and epistemic uncertainty in a single forward pass.”
Kano, K., Mori, Y., Higashiura, K., Hossain, T., Katayama, S., Urano, K., Yonezawa, T., & Kawaguchi, N. (2024) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3675094.3678447 · Cited by 4
“Choi et al. used Mixture Density Network (MDN) for object detection to estimate aleatoric and epistemic uncertainty on localization and classification in a single forward pass [3].”
Optimizing Object Detection via Metric-driven Training Data Selection
Zhou, C., Guo, Y., Lv, Q., & Ji, Y. (2024) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00730 · Cited by 4
“Parallel to domain adaptation, active machine learning has emerged as a paradigm to optimize the training process by selectively querying the most informative data points [2, 16].”
Joint Semi-Supervised and Active Learning via 3D Consistency for 3D Object Detection
Hwang, S., Kim, S., Kim, Y., & Kum, D. (2023) · IEEE International Conference on Robotics and Automation · doi:10.1109/icra48891.2023.10160433 · Cited by 7
“The uncertainty-based approaches [10]−[16], on the other hand, focus on selecting samples with the highest uncertainty in the model prediction by measuring the entropy [34], [11] or probability of class [10].”
BackgroundClassification Committee for Active Deep Object Detection
Zhao, L., Li, B., Jiang, J.-X., & Wei, X. (2023) · IEEE transactions on multimedia · doi:10.1109/tmm.2024.3521778 · Cited by 4
“The proposed method is consistent with mainstream active learning meth-ods [38], [50], [51].”
InfluentialDeep active learning: A reality check
Gashi, E., Deng, J., & Elezi, I. (2025) · Pattern Recognition Letters · doi:10.1016/j.patrec.2025.07.017 · Cited by 3
Self-citationMIME: Minority Inclusion for Majority Group Enhancement of AI Performance
Chari, P., Ba, Y., Athreya, S., & Kadambi, A. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19778-9_19 · Cited by 3
Uncertainty Quantification in Table Structure Recognition
Ajayi, K., Zhang, L., He, Y., & Wu, J. (2024) · IEEE International Conference on Information Reuse and Integration (IRI) · doi:10.1109/iri62200.2024.00014 · Cited by 3
“We also compare our method against an active learning model proposed by Choi et al. [22].”
ResultLiu, P., Huang, B., Jin, T., & Long, H. (2025) · Remote Sensing · doi:10.3390/rs17071155 · Cited by 3
Yang, J., Wu, J., Xiao, S., Wen, J., Meng, Q., Lu, W. B., & Gao, X. (2025) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2025.3584808 · Cited by 3
“Additionally, considering the prevalent adoption of dual-branch architectures (consisting of classification and regression modules) in detection frameworks [26], achieving information alignment between these complementary branches continues to pose substantial technical challenges.”
Wang, J., Chen, L. F., & Wang, X. (2023) · Wireless networks · doi:10.1007/978-3-031-32397-3_15 · Cited by 3
Select Informative Samples for Night-Time Vehicle Detection Benchmark in Urban Scenes
Wang, X., Tu, X., Al-Hassani, B., Lin, C., & Xu, X. (2023) · Remote Sensing · doi:10.3390/rs15174310 · Cited by 3
Schmidt, S., Stappen, L., Schwinn, L., & Günnemann, S. (2024) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2024.3444670 · Cited by 7
“Latter research directions employ Gaussian mixture models to asses uncertainty [13], [14].”
MethodIntrospective Perception for Long-Term Aerial Telemanipulation With Virtual Reality
Lee, J., Balachandran, R., Kondak, K., Coelho, A., de Stefano, M., Humt, M., Feng, J., Asfour, T., & Triebel, R. (2024) · IEEE Transactions on Field Robotics · doi:10.1109/tfr.2024.3494719 · Cited by 3
“Recent works can also be found on active learning for DNN based object detectors [54], [55], where the focus is on adaptations of active learning to existing object detection frameworks.”
Inconsistency-Based Active Learning for LiDAR Object Detection
Rivera, E., Stratil, L., & Lienkamp, M. (2025) · IEEE Intelligent Vehicles Symposium · doi:10.1109/iv64158.2025.11097777 · Cited by 2
“For example, Choi[14] proposes using mixture density networks to estimate a probabilistic distribution for each localization and classification head’s output.”
Rarity-Aware Stratified Active Learning for Class-Imbalanced Industrial Object Detection
Benhafid, Z., & Selouani, S. (2026) · Applied Sciences · doi:10.3390/app16031236 · Cited by 2
“Probabilistic modeling approaches explicitly estimate uncertainty for bounding box regression and class prediction within a single model [32].”
Active learning for industrial defect detection: a study on hybrid sampling strategies
Gonzalez, D. G., Nascimento, R. C., Rocha, C. D., Silva, M. F., Filipe, V. M. D. J., Rocha, L. F., Magalhães, L., & Cunha, A. M. T. D. S. (2026) · The International Journal of Advanced Manufacturing Technology · doi:10.1007/s00170-025-17378-7 · Cited by 2
Feature Augmentation Based on Pixel-Wise Attention for Rail Defect Detection
Li, H., Li, H., Hou, Z., Song, H., Liu, J., & Dai, P. (2022) · Applied Sciences · doi:10.3390/app12168006 · Cited by 2
“As a comparison, AL-MDN and DyHead also outperform traditional one-stage and two-stage baselines, and are competitive with our proposed method.”
MethodResultVideo logo detection by Deep-Transfer Active Learning
Su, H., Qiu, G., & Guangdong Key Lab for Intelligent Information Processing, C. (2022) · Discrete and Continuous Dynamical Systems - S · doi:10.3934/dcdss.2022181 · Cited by 2
Na, J., & De-Silva, V. (2023) · 2023 IEEE IAS Global Conference on Emerging Technologies (GlobConET) · doi:10.1109/globconet56651.2023.10149954
“However, [19], [20] has shown that the data informativeness for the localization component is also crucial to the performance of an active learning algorithm.”
BackgroundMethodInconsistency-based active learning with adaptive pseudo-labeling for fish species identification
Nabi, M. M., Shah, C., Alaba, S. Y., Caillouet, R., Prior, J. H., Campbell, M. D., Wallace, F., Ball, J. E., & Moorhead, R. J., II (2024) · SPIE Ocean Sensing and Monitoring · doi:10.1117/12.3013336 · Cited by 2
“32 In Table 1, we conduct a total of five cycles, with cycle 0 representing the initial supervised learning model, starting with 3,000 samples.”
MethodObject Recognition Consistency in Regression for Active Detection
Jing, M., Ou, Z., Wang, H.-X., Li, J., & Zhao, Z. (2024) · Machine Vision and Applications · doi:10.1007/s00138-024-01604-5 · Cited by 3
“The second type of methods attempts to calculate the localization uncertainty of the objects, such as AL-MDN [21].”
InfluentialTSceneJAL: Joint Active Learning of Traffic Scenes for 3D Object Detection
Lei, C., Peng, W.-Y., Zhou, G., Zhang, M.-Y., Hao, Q., Ji, C., & Xu, C.-Z. (2024) · IEEE transactions on intelligent transportation systems (Print) · doi:10.1109/tits.2025.3553170 · Cited by 2
“While several metrics, like statistical complexity [20], [21] and perception uncertainty [10], [22], [23], have been proposed to gauge scene complexity.”
HiCAL: Hierarchical Consistency-Based Active Learning for Drone-View Object Detection
Zhang, X., Liu, Y., Li, A., Han, K., Zhao, Q., & Wu, J. (2025) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2025.3576980 · Cited by 3
“Common uncertainty measurement methods include Monte Carlo uncertainty sampling [12], entropy [13], similarity sampling [14], and some other training-embedded approaches [39], [40], [41].”
Shin, J. W., & Kim, H. (2023) · International Conference on Electronics, Information, and Communication (ICEIC) · doi:10.1109/iceic57457.2023.10049902 · Cited by 2
“However, the performance of each task is highly dependent on the training data, and the use of limited data hinders the performance improvement of vision tasks [5].”
BackgroundSiamese Network Representation for Active Learning
Li, J., Du, Y., & Du, L. (2023) · Proceedings - International Conference on Image Processing · doi:10.1109/icip49359.2023.10222798 · Cited by 2
“Active learning can be categorized in uncertainty-based methods[2, 3], query-by-committee methods[4, 5, 6], representation-based methods[7, 8, 9, 10], etc.”
BackgroundHierarchical Uncertainty Aggregation and Emphasis Loss for Active Learning in Object Detection
Nguyen, T. T., Nguyen, K. D., Nguyen, T., Nguyen, T. C., Nguyen, A., & Kim, K. (2023) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata59044.2023.10386534 · Cited by 2
“Diverging from these approaches, active learning [18], [19], [20], [21], [22] emerges as a promising solution.”
BackgroundBucketized Active Sampling for learning ACOPF
Klamkin, M., Tanneau, M., Mak, T. W. K., & Hentenryck, P. V. (2024) · Electric Power Systems Research · doi:10.1016/j.epsr.2024.110697 · Cited by 3
“Acquisition functions α (or scoring functions [23]) are commonly used in active sampling applications to select the best k samples from D for the next training cycle.”
BackgroundWang, M., Liu, G., & Liu, H. (2024) · IEEE Access · doi:10.1109/access.2024.3372595 · Cited by 1
“Despite the success of these object detection methods [8], [29], certain limitations persist.”
BackgroundMethodInfluentialMurel, J., & Smith, D. A. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-70543-4_18 · Cited by 1
Hierarchical evidence aggregation in two dimensions for active water surface object detection
Zhong, W., Hu, H., Wang, Y., Li, L., Han, T., Li, C., & Zan, P. (2024) · The Visual Computer · doi:10.1007/s00371-024-03685-9 · Cited by 1
“From different perspectives, AL can be broadly categorized into diversity-based [21, 22], uncertainty-based [5, 23–26] and hybrid-based [27–29].”
SSE: Multimodal Semantic Data Selection and Enrichment for Industrial-scale Data Assimilation
Shen, M., Chang, N., Liu, S., & Álvarez, J. M. (2025) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3690624.3709417 · Cited by 1
“Crucially, we quantitatively show that the now semantically expanded dataset is able to improve rare class performance with even fewer rare but semantically important objects, demonstrating the strong advantage semantics imbue.”
Kun, 唐. 昆. T., Zhao, 黄. H., Yongjian, 朱. Z., Hang, 张. Z., Gang, 曾. Z., Hongchao, 肖. X., Xiaojie, 周. Z., Weidong, 唐. T., Mingjun, 张. Z., & Cong, 毛. M. (2025) · Chinese Journal of Lasers · doi:10.3788/cjl241344 · Cited by 1
Enhanced uncertainty sampling with category information for improved active learning
Wang, X., Zhang, B., Wang, F., Bao, T., Lu, Z., & Bao, J. (2025) · PLoS ONE · doi:10.1371/journal.pone.0327694 · Cited by 1
“In [18], a deep active learning method designed for object detection is proposed.”
Table Detection with Active Learning
Gautam, S., Purohit, N., & Harit, G. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-04624-6_8 · Cited by 1
“Although AL has been successfully applied to various object detection tasks [28], [11], [18], [13], [4], [40], [41], [5], [39], its application to particularly table detection remains unexplored.”
Efficient Iterative Semi-supervised Object Detection Training for Counting of Fruits and Vegetables
Bogomasov, K., & Conrad, S. (2023) · International Conference on Advances in Artificial Intelligence (ICAAI) · doi:10.1145/3633598.3633619 · Cited by 1
“On this basis, a couple of AL-publications for OD have been introduced [6], [7], [9], [5], [15].”
BackgroundEnhancing Pseudo-Labeling Performance in Object Detection Using Gaussian Mixture Modeled Uncertainty
Lee, S., & Kim, H. (2024) · International Conference on Electronics, Information, and Communication (ICEIC) · doi:10.1109/iceic61013.2024.10457103 · Cited by 1
“GMM [9], [10] represents the network’s output y for an input image x with parameters consisting of mean( µ ), variance( Σ ), and weights( π ) as follows: where k is the number of model components used in GMM, and the extracted parameters are used to redefine the loss function.”
BackgroundMethodActive Batch Sampling for Multi-label Classification with Binary User Feedback
Goswami, D., & Chakraborty, S. (2024) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv57701.2024.00252 · Cited by 1
“Batch mode AL has been successfully used in a variety of computer vision applications, such as image recognition [37], semantic segmentation [1], object detection [3] and image regression [19] among others.”
MethodActive Learning Method Based on Pseudo-labeling
Hou, X., & Wang, C. (2024) · International Conference on Robotics, Control and Automation · doi:10.1109/icrca60878.2024.10649264 · Cited by 2
“Choi et al. [13] proposed to use a mixed density network to estimate the probability distribution of the output of each positioning head and classification head, fully evaluate the arbitrary uncertainty and cognitive uncertainty, and eliminate the interference of noise.”
BackgroundOSA-PD: Open-Set Active Learning Exploiting Predictive Discrepancy
Chen, Z., Han, P., Jiang, F., Si, J., & Xiong, L. (2025) · IEEE Signal Processing Letters · doi:10.1109/lsp.2025.3537398 · Cited by 1
“For instance, Entropy [13] utilizes information entropy to assess uncertainty, a concept further generalized in subsequent studies such as [14], [15].”
Enhancing weakly supervised 3D medical image segmentation through probabilistic-aware learning
Jiang, R., Fan, Z., Wu, J., Zhu, L., Huang, X., Wang, T., Huang, H., & Xu, M. (2026) · Pattern Recognition · doi:10.1016/j.patcog.2026.113973 · Cited by 5
“Label Generation”
Semi-Automatic Video Frame Annotation for Construction Equipment Automation Using Scale-Models
Borngrund, C., Hammarkvist, T., Bodin, U., & Sandin, F. (2021) · Annual Conference of the IEEE Industrial Electronics Society (IECON) · doi:10.1109/iecon48115.2021.9589255 · Cited by 1
“Active learning has also been used to improve the convergence of object detectors [38], [39], with promising results.”
BackgroundSource-Free Unsupervised Domain Adaptation in Imbalanced Datasets
Sun, W., Chen, Q., Wang, C., Ye, D., & Chen, C. (2022) · International Conference on Data Science and Information Technology (DSIT) · doi:10.1109/dsit55514.2022.9943839 · Cited by 2
“It is worth noting that although entropy can be used in active learning [3, 15, 16] to measure the uncertainty of the sample [22], it differs greatly from SFUDA and DEB.”
BackgroundMulti-granularity active learning based on the three-way decision
Wu, X., & Thitipong, T. (2023) · International Journal of Advances in Intelligent Informatics · doi:10.26555/ijain.v9i2.1036
“Active learning can reduce labeling costs by selecting the most informative samples of the dataset [14], [15].”
BackgroundQuery-based Hard-Image Retrieval for Object Detection at Test Time
Ayers, E. W., Sadeghi, J., Redford, J., Mueller, R., & Dokania, P. (2022) · AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v37i12.26717 · Cited by 3
“…objective of these works was to either perform active learning mainly to reduce the labelling and training cost (Gal, Is-lam, and Ghahramani 2017; Choi et al. 2021; Roy, Unmesh, and Namboodiri 2018; Aghdam et al. 2019; Haussmann et al. 2020; Yu et al. 2022; Feng et al. 2019), or to improve the…”
BackgroundEpistemic uncertainty estimation for object detection in open-set conditions
Miller, D. (2021) · Queensland University of Technology · doi:10.5204/thesis.eprints.213588 · Cited by 1
Metadata
- Authors
- Choi, J., Elezi, I., Lee, H.-J., Farabet, C., & Alvarez, J. M.
- Affiliations
- Seoul National University (KR)
- Ca' Foscari University of Venice (IT)
- NVIDIA
- Venue
- 2021 IEEE/CVF International Conference on Computer Vision (ICCV)
- Publisher
- IEEE
- Published
- 2021-10
- Pages
- 10244-10253
- OpenAlex
- W3145121851
- Cited by
- 135
- References
- 49
- Citations by year
- 2026: 15 · 2025: 27 · 2024: 36 · 2023: 36 · 2022: 17 · 2021: 4
- FWCI
- 10.74
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Algorithms
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Keywords
- active learning (machine learning), object detection, mixture density networks, uncertainty estimation, aleatoric uncertainty, epistemic uncertainty, annotation cost reduction
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
