Scalable Active Learning for Object Detection
Open accessAbstract
Deep Neural Networks trained in a fully supervised fashion are the dominant technology in perception-based autonomous driving systems. While collecting large amounts of unlabeled data is already a major undertaking, only a subset of it can be labeled by humans due to the effort needed for high-quality annotation. Therefore, finding the right data to label has become a key challenge. Active learning is a powerful technique to improve data efficiency for supervised learning methods, as it aims at selecting the smallest possible training set to reach a required performance. We have built a scalable production system for active learning in the domain of autonomous driving. In this paper, we describe the resulting high-level design, sketch some of the challenges and their solutions, present our current results at scale, and briefly describe the open problems and future directions.
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
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?
Mutual information with a maximum seems to offer the best trade-off between mAP and samples count.
“The result of the extensive evaluation was that mutual information with a maximum seems to offer the best trade-off between mAP and samples count.”
From Ten Years of Active Learning Techniques and Object Detection: A Systematic Review (2023) · cites this paper as [36]
The Scoring functions section says that MI using Max provides the best trade-off when labeling cost is taken into account. But Table I compares wMAP with the number of bounding boxes produced, not the number of image samples, so the claim's “samples count” does not match the reported measure. (LLM, conf=0.86)
Experimental — accuracy not yet measuredAsia (China), North America (United States of America), and Europe (Germany) are producing the main advances in active learning for object detection research.
“An analysis of the research groups or technological companies working in AL for OD showed that Asia (China), North America (United States of America), and Europe (Germany) are producing the main advances in this research field (e.g., [5,36,55]), as can be seen in Figure 5.”
From Ten Years of Active Learning Techniques and Object Detection: A Systematic Review (2023) · cites this paper as [36]
The use of active learning in object detection needs more attention.
“while its use in other tasks, such as OD, needs more attention [36].”
From Ten Years of Active Learning Techniques and Object Detection: A Systematic Review (2023) · cites this paper as [36]
In hybrid methods combining uncertainty- and diversity-based methods, diversity can prevent instances from overlapping or becoming redundant in queries, producing better sampling results, or can overcome noisy labels.
“For example, in the combination of uncertainty- and diversity-based methods, diversity can prevent instances from overlapping or becoming redundant in queries producing better sampling results [36,67], or can overcome noisy labels [61].”
From Ten Years of Active Learning Techniques and Object Detection: A Systematic Review (2023) · cites this paper as [36]
The most advanced companies and research centres in AI are interested in continuing research around active learning and exploring its potential benefits further.
“On the other hand, the most advanced companies and research centres in the field of AI are also interested in continuing research around AL and exploring its potential benefits further [14,36,50,59,78].”
From Ten Years of Active Learning Techniques and Object Detection: A Systematic Review (2023) · cites this paper as [36]
Proposed methods do not always fully explore their potential by focusing on an exclusive criterion when considering several complementary sampling strategies simultaneously is possible.
“However, the proposed methods do not always fully explore their potential by focusing on an exclusive criteria-among the previously referenced ones-when considering several complementary sampling strategies simultaneously is possible [36,37,44,45,57,61,67].”
From Ten Years of Active Learning Techniques and Object Detection: A Systematic Review (2023) · cites this paper as [36]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
You Only Look Once: Unified, Real-Time Object Detection
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.91 · Cited by 53,803 · see citing works
(2016) · Cited by 31,590 · see citing works
(1967) · Cited by 23,719 · see citing works
(2007) · doi:10.5555/1283383.1283494 · Cited by 7,364 · see citing works
Dropout as a bayesian approximation: Representing model uncertainty in deep learning
Gal, Y., & Ghahramani, Z. (2015) · arXiv · doi:10.48550/arxiv.1506.02142 · Cited by 6,471 · see citing works
Pati, Y., Rezaiifar, R., & Krishnaprasad, P. (1993) · Proceedings of 27th Asilomar Conference on Signals, Systems and Computers · doi:10.1109/acssc.1993.342465 · Cited by 4,928 · see citing works
(2009) · Minds at UW (University of Wisconsin) · Cited by 4,791 · see citing works
The Power of Ensembles for Active Learning in Image Classification
Beluch, W. H., Genewein, T., Nurnberger, A., & Kohler, J. M. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00976 · Cited by 655 · see citing works
Active Learning for Convolutional Neural Networks: A Core-Set Approach
Şener, O., & Savarese, S. (2018) · International Conference on Learning Representations · Cited by 460 · see citing works
Deep Batch Active Learning by Diverse, Uncertain Gradient Lower BoundsPreprint
Ash, J. T., Zhang, C., Krishnamurthy, A., Langford, J. C., & Agarwal, A. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1906.03671 · Cited by 366 · see citing works
Bias-Reduced Uncertainty Estimation for Deep Neural ClassifiersPreprint
Geifman, Y., Uziel, G., & El-Yaniv, R. (2018) · International Conference on Learning Representations · doi:10.48550/arxiv.1805.08206 · Cited by 212 · see citing works
Active Learning for Deep Detection Neural Networks
Aghdam, H. H., Gonzalez-Garcia, A., Lopez, A., & Weijer, J. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00377 · Cited by 148 · cited 1× · see citing works
“[15] propose an image-level scoring process to rank unlabeled images for”
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 · see citing works
Deep active learning for object detection
Roy, S. D., Unmesh, A., & Namboodiri, V. P. (2018) · Cited by 71 · see citing works
Uncertainty Estimation via Stochastic Batch Normalization
Atanov, A., Ashukha, A., Molchanov, D., Neklyudov, K., & Vetrov, D. (2019) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-22796-8_28 · Cited by 50 · see citing works
Uncertainty-Based Active Learning via Sparse Modeling for Image Classification
Wang, G., Hwang, J., Rose, C. S., et al. (2018) · IEEE Transactions on Image Processing · doi:10.1109/tip.2018.2867913 · Cited by 40 · see citing works
Large-Scale Visual Active Learning with Deep Probabilistic EnsemblesPreprint
Chitta, K., Álvarez, J. M., & Lesnikowski, A. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.03575 · Cited by 29 · see citing works
Self-citationAn Adaptive Supervision Framework for Active Learning in Object Detection
Desai, S. V., Chandra, A. L., Guo, W., Ninomiya, S., & Balasubramanian, V. N. (2019) · arXiv · doi:10.48550/arxiv.1908.02454 · Cited by 29 · see citing works
Active Learning for Deep Detection Neural NetworksPreprint
Aghdam, H. H., González-García, A., van de Weijer, J., & López, A. M. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1911.09168 · Cited by 17 · see citing works
Adaptive Semantic Segmentation with a Strategic Curriculum of Proxy LabelsPreprint
Chitta, K., Feng, J., & Hebert, M. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.03542 · Cited by 12 · see citing works
Self-citationLess is More: An Exploration of Data Redundancy with Active Dataset Subsampling.Preprint
Chitta, K., Álvarez, J. M., Haußmann, E., & Farabet, C. (2019) · arXiv (Cornell University) · Cited by 12 · see citing works
Self-citation
Flagged first, then by citations
Research works printed but not linked: 2
- B. Lakshminarayanan, A. Pritzel, and C. Blundell, "Simple and scalable predictive uncertainty estimation using deep ensembles," in NIPS, 2017.
- Y. Gal, "Uncertainty in deep learning," Ph.D. dissertation, University of Cambridge, 2016.
Cited by
Sorted by citations
Active Learning for Deep Object Detection via Probabilistic Modeling
Choi, J., Elezi, I., Lee, H.-J., Farabet, C., & Alvarez, J. M. (2021) · 2021 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.01010 · Cited by 135 · see citing works
“These methods reach good results but come with several limitations [11, 16].”
BackgroundMethodInfluentialSelf-citationTen 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 · see citing works
Not All Labels Are Equal: Rationalizing The Labeling Costs for Training Object Detection
Elezi, I., Yu, Z., Anandkumar, A., Leal-Taixe, L., & Alvarez, J. M. (2021) · arXiv · Cited by 50 · see citing works
“However, the state-of-the-art approaches are based on the uncertainty [4, 5, 8, 26].”
BackgroundSelf-citation
- Authors
- Haussmann, E., Fenzi, M., Chitta, K., Ivanecky, J., Xu, H., Roy, D., Mittel, A., Koumchatzky, N., Farabet, C., & Alvarez, J. M.
- Affiliations
- Nvidia (United States) (US)
- Max Planck Institute for Intelligent Systems (DE)
- ORCID
- Chitta, Kashyap
- Venue
- 2020 IEEE Intelligent Vehicles Symposium (IV)
- Publisher
- IEEE
- Published
- 2020-10-19
- Pages
- 1430-1435
- ISSN
- 1931-0587 (ISSN-L)
- OpenAlex
- W3015770003
- Cited by
- 110
- References
- 21
- Citations by year
- 2026: 2 · 2025: 14 · 2024: 25 · 2023: 25 · 2022: 19 · 2021: 20 · 2020: 5
- FWCI
- 6.59
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Intelligent Vehicles Symposium (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- IV
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Algorithms
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- active learning (machine learning), object detection, autonomous driving, data annotation, data efficiency, annotation cost, training data selection
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
