The DOI matches its Crossref record; 25 of the 30 references it cites were found in a registry or an index.
Active Learning for Deep Detection Neural Networks
Open accessAbstract
The cost of drawing object bounding boxes (i.e. labeling) for millions of images is prohibitively high. For instance, labeling pedestrians in a regular urban image could take 35 seconds on average. Active learning aims to reduce the cost of labeling by selecting only those images that are informative to improve the detection network accuracy. In this paper, we propose a method to perform active learning of object detectors based on convolutional neural networks. We propose a new image-level scoring process to rank unlabeled images for their automatic selection, which clearly outperforms classical scores. The proposed method can be applied to videos and sets of still images. In the former case, temporal selection rules can complement our scoring process. As a relevant use case, we extensively study the performance of our method on the task of pedestrian detection. Overall, the experiments show that the proposed method performs better than random selection.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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?
In the citing paper
“Multiple alternatives include the aggregation of information from multiple sources such as: pixel-level, box-level [45] or image-level [44,47], which have been useful for both uncertainty-based and diversity-based methods”
From Ten Years of Active Learning Techniques and Object Detection: A Systematic Review (2023) · cites this paper as [47]
Citations
Cites
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
Rethinking Atrous Convolution for Semantic Image Segmentation
Chen, L.-C., Papandreou, G., Schroff, F., & Adam, H. (2017) · arXiv · doi:10.48550/arxiv.1706.05587 · Cited by 7,517 · cited 1×
“To this end, we follow [3] and resize the logits spatially using the bilinear upsampling.”
MethodSqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size
Iandola, F. N., Han, S., Moskewicz, M. W., Ashraf, K., Dally, W. J., & Keutzer, K. (2016) · arXiv · doi:10.48550/arxiv.1602.07360 · Cited by 6,201 · cited 1×
“er must have the same size as the image. To this end, we follow [3] and resize the logits spatially using the bilinear upsampling. Each Fire Residual module follows the same architecture indicated in [12]. Also, each downsampling layer is composed of convolution and pooling layers which are applied in parallel and con…”
MethodSimple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
Lakshminarayanan, B., Pritzel, A., & Blundell, C. (2016) · arXiv · doi:10.48550/arxiv.1612.01474 · Cited by 3,666 · cited 1×
“[14] and Gal et al .”
Pedestrian Detection: An Evaluation of the State of the Art
Dollár, P., Wojek, C., Schiele, B., & Perona, P. (2011) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2011.155 · Cited by 3,232 · cited 2×
“CaltechPed[11] • 640x480 resolution • 51363 Images • 20062 Boxes • Videos”
Deep Visual Domain Adaptation: A Survey
Wang, M., & Deng, W.-H. (2018) · Neurocomputing · doi:10.48550/arxiv.1802.03601 · Cited by 2,451 · cited 1×
“In other words, there is a domain shift [19, 25, 28, 27] between Xl and Xu that makes the active learning procedure more challenging.”
BackgroundTorralba, A., & Efros, A. A. (2011) · CVPR 2011 · doi:10.1109/cvpr.2011.5995347 · Cited by 2,440 · cited 1×
“In other words, there is a domain shift [19, 25, 28, 27] between Xl and Xu that makes the active learning procedure more challenging.”
BackgroundChandola, V., Banerjee, A., & Kumar, V. (2013) · The Veterinary Record · doi:10.1007/978-1-4899-7687-1_916 · Cited by 2,312 · cited 1×
“[2] computed the marginal score [23] of candidate bounding boxes and integrate them using different merging functions.”
Deep visual domain adaptation: A survey
Wang, M., & Deng, W. (2018) · Neurocomputing · doi:10.1016/j.neucom.2018.05.083 · Cited by 2,209
ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation
Paszke, A., Chaurasia, A., Kim, S., & Culurciello, E. (2016) · arXiv · doi:10.48550/arxiv.1606.02147 · Cited by 1,826 · cited 1×
“Also, each downsampling layer is composed of convolution and pooling layers which are applied in parallel and concatenated at the output [17].”
BackgroundDataset Shift in Machine Learning
(2008) · doi:10.7551/mitpress/9780262170055.001.0001 · Cited by 1,553 · cited 2×
“In other words, there is a domain shift [19, 25, 28, 27] between Xl and Xu that makes the active learning procedure more challenging.”
BackgroundCityPersons: A Diverse Dataset for Pedestrian Detection
Zhang, S., Benenson, R., & Schiele, B. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.474 · Cited by 1,061
Deep Bayesian Active Learning with Image Data
Gal, Y., Islam, R., & Ghahramani, Z. (2017) · arXiv · doi:10.48550/arxiv.1703.02910 · Cited by 900 · cited 12×
“This score has been previously used by [11] and [8] in the task of image classification.”
MethodInfluentialBayesian Convolutional Neural Networks with Bernoulli Approximate Variational InferencePreprint
Gal, Y., & Ghahramani, Z. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1506.02158 · Cited by 777 · cited 1×
“[7, 6, 15] add a prior on the weights of image classification neural networks, sampling the weights from the dropout distribution at each evaluation.”
BackgroundBayesian Active Learning for Classification and Preference Learning
Houlsby, N., Huszár, F., Ghahramani, Z., & Lengyel, M. (2011) · arXiv · doi:10.48550/arxiv.1112.5745 · Cited by 507
Deep Bayesian Active Learning with Image Data
Ghahramani, Z., Gal, Y., & Islam, R. (2017) · PMLR · doi:10.17863/cam.11070 · Cited by 392
BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning
Yu, F., Chen, H., Wang, X., Xian, W., Chen, Y., Liu, F., Madhavan, V., & Darrell, T. (2018) · arXiv · doi:10.48550/arxiv.1805.04687 · Cited by 204
Deep Active Learning over the Long TailPreprint
Geifman, Y., & El‐Yaniv, R. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1711.00941 · Cited by 162
Domain Adaptation of Deformable Part-Based Models
Xu, J.-L., Ramos, S., Vázquez, D., & López, A. (2014) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2014.2327973 · Cited by 153 · cited 1×
“In other words, there is a domain shift [19, 25, 28, 27] between X l and X u that makes the active learning procedure more challenging.”
BackgroundActive Learning for Deep Object DetectionPreprint
Brust, C., Käding, C., & Denzler, J. (2018) · VISIGRAPP · doi:10.48550/arxiv.1809.09875 · Cited by 148
Localization-Aware Active Learning for Object Detection
Kao, C.-C., Lee, T.-Y., Sen, P., & Liu, M.-Y. (2018) · Asian Conference on Computer Vision · doi:10.48550/arxiv.1801.05124 · Cited by 146 · cited 1×
“Active Learning for Object Detection (Kao et al.[5])”
BackgroundDropout Inference in Bayesian Neural Networks with Alpha-divergences
Li, Y., & Gal, Y. (2017) · Oxford University Research Archive (ORA) (University of Oxford) · doi:10.48550/arxiv.1703.02914 · Cited by 145 · cited 1×
“[7, 6, 15] add a prior on the weights of image classification neural networks, sampling the weights from the dropout distribution at each evaluation.”
BackgroundA Convex Optimization Framework for Active Learning
Elhamifar, E., Sapiro, G., Yang, A. Y., & Sastry, S. (2013) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2013.33 · Cited by 134 · cited 1×
“[5] formulated the selection as a convex optimization problem taking into account the similarity of selected samples in the feature space as well as their informativeness score.”
BackgroundActive Learning for Deep Object Detection
Brust, C.-A., Käding, C., & Denzler, J. (2019) · Proceedings of the 14th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications · doi:10.5220/0007248601810190 · Cited by 78
What Do I Annotate Next? An Empirical Study of Active Learning for Action Localization
Heilbron, F. C., Lee, J., Jin, H., & Ghanem, B. (2018) · Lecture notes in computer science · doi:10.1007/978-3-030-01252-6_13 · Cited by 43 · cited 1×
“[10] for action localization.”
Flagged first, then by citations
5 of the 30 references couldn’t be matched to a known paper.
Research works printed but not linked: 7
- Abramson, 2005, SEmi-automatic VIsuaL LEarning (SEVILLE): a tutorial on active learning for visual object recognition, CVPR
- 10.1007/978-3-030-20876-9_32 · DOI · Paper page
- Paul, 2014, ICRA
- Sener, 2018, Active learning for convolutional neural networks: A core-set approach, ICML
- Settles, 2010, Active learning literature survey
- 10.2200/S00429ED1V01Y201207AIM018 · DOI · Paper page
- 10.1007/s00138-011-0388-y · DOI
Cited by
Sorted by citations
A Survey of Deep Active Learning
Ren, P., Xiao, Y., Chang, X., Huang, P.-Y., Li, Z., Gupta, B. B., Chen, X., & Wang, X. (2022) · ACM Computing Surveys · doi:10.1145/3472291 · Cited by 1,169
“tion so as to obtain high-quality query samples. Fig.3 presents a comparison between traditional AL, RAL and DRAL pipelines. On the other hand, Active-iNAS [50] notices that most previous DAL methods [3, 5, 89] assume that a suitable DL model has been designed for the current task, meaning that their primary focus is…”
BackgroundMethodInfluentialArtificial intelligence in recommender systems
Zhang, Q., Lü, J., & Jin, Y. (2020) · Complex & Intelligent Systems · doi:10.1007/s40747-020-00212-w · Cited by 519
“A system with an active learning strategy may query users to provide labels for unlabeled instances [46].”
BackgroundLocalizing Objects with Self-Supervised Transformers and no LabelsPreprint
Siméoni, O., Puy, G., Vo, H. V., Roburin, S., Gidaris, S., Bursuc, A., Pérez, P., Marlet, R., & Ponce, J. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2109.14279 · Cited by 190
“Various approaches have been proposed to reduce these costs, such as semi-supervision [41], weak supervision [52], active-learning [3] and self-supervision [26] with task fine-tuning.”
BackgroundFernandes, D., Silva, A., Névoa, R., Simões, C., Gonzalez, D. G., Guevara, M. A., Nováis, P., Monteiro, J. L., & Melo‐Pinto, P. (2020) · Information Fusion · doi:10.1016/j.inffus.2020.11.002 · Cited by 283
Wang, Y., Shen, X., Yuan, Y., Du, Y., Li, M., Hu, S. X., Crowley, J. L., & Vaufreydaz, D. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3305122 · Cited by 122
“Researchers have attempted to overcome this barrier using active learning [1, 47], semi-supervised learning [8, 36], and weakly-supervised learning [26, 43] with limited results.”
Background
Metadata
- Authors
- Aghdam, H. H., Gonzalez-Garcia, A., Lopez, A., & Weijer, J.
- Affiliations
- Universitat Autònoma de Barcelona (ES)
- ORCID
- Aghdam, Hamed H.
- Venue
- 2019 IEEE/CVF International Conference on Computer Vision (ICCV)
- Publisher
- IEEE
- Published
- 2019-10
- Pages
- 3671-3679
- OpenAlex
- W2984303785
- Cited by
- 148
- References
- 30
- Citations by year
- 2026: 9 · 2025: 14 · 2024: 24 · 2023: 31 · 2022: 32 · 2021: 28 · 2020: 9 · 2019: 1
- FWCI
- 6.70
- Type
- conference paper
- Language
- en
- Open access
- green
- ddd.uab.cat
- Licence
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- Dipòsit Digital de Documents de la UAB (Universitat Autònoma de Barcelona) (submittedVersion, public-domain)
- 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, pedestrian detection, convolutional neural networks, video frame selection, 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
