Pith. sign in

REVIEW 3 major objections 5 minor 33 references

Exploring Spatial Diversity for Region-based Active Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Spread-out regions reach 95% of full supervision with 9% of pixels

desk verdict The empirical result is solid, but the proposed spatial-diversity mechanism likely reduces to image-level coverage; still a useful, well-ablated contribution that deserves peer review. read the letter →

arxiv 2507.17367 v1 pith:TBZUJ5OT submitted 2025-07-23 cs.CV

Lile Cai , Xun Xu , Lining Zhang , Chuan-Sheng Foo This is my paper · ORCID
classification cs.CV
keywords activelearningsemanticsegmentationregion-basedspatialdiversityentropysamplingmax-minoptimizationlabelefficiency
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that when active learning labels image regions instead of whole images, selected regions should be spread out in image space as well as uncertain. Nearby regions in a segmentation image tend to share labels, so a batch chosen only by uncertainty is redundant. The authors add a spatial-diversity term, computed purely from region coordinates, to a max-min selection objective and solve it greedily. On Cityscapes and PASCAL VOC 2012, the resulting Entropy+Spatial method reaches about 95% of the accuracy of a fully supervised model while labeling only 8.4% and 5.9% of pixels, respectively.

What carries the argument

The load-bearing object is the piece-wise constant spatial distance $d_s(x_i,x_j)$, which assigns value $a$ to regions within distance $\tau$ in the same image, $b$ to regions farther apart in the same image, and $c$ to regions from different images, with $c \ge b \ge a > 0$ and $b \le 2a$ so it is a metric. This distance is inserted into a max-min objective $\max_{B_t} \min[\lambda_u u(x_k) + d(x_i,x_j)]$ in which the unary term is per-region entropy and the pairwise term is diversity. A greedy algorithm selects, at each step, the region maximizing the minimum over all already selected regions of entropy plus distance, which scales to the roughly $10^5$ regions in benchmark segmentation datasets.

What would settle it

A controlled experiment that permutes region coordinates while keeping the entropy values fixed should make the spatial term's benefit disappear if the benefit comes from the coordinate-label correlation; if Entropy+Spatial still beats Entropy after such permutation, the spatial term is doing something else. Concretely, run the paper's Cityscapes protocol with region locations randomly shuffled across images and compare mIoU curves.

Watch

Extended reading notes

Core claim

The central claim is that spatial diversity is a useful objective for region-based active learning, and can replace feature-space diversity for semantic segmentation. The paper shows that combining per-region entropy with a piece-wise spatial distance, which penalizes selecting nearby regions while treating far-apart and cross-image regions equally, improves over entropy alone and over feature-diversity baselines such as core-set and entropy-plus-feature in the reported experiments. Because the spatial distance depends only on region coordinates, it remains reliable when the feature extractor is poorly trained or when feature distances concentrate in high dimensions. The authors also report that the method outperforms a reinforcement-learning region selector at a smaller labeling budget.

Load-bearing premise

The central assumption is that image coordinates are a reliable stand-in for label diversity: regions that are far apart are assumed to contain different semantic categories, so spreading selections across the image removes redundancy.

Editorial extensions

If this is right

  • Region-based active learning can reach near-fully-supervised accuracy with a few percent of labeled pixels, so annotation budgets for semantic segmentation can shrink by an order of magnitude.
  • Spatial diversity removes the need for a trained feature extractor during selection, making each active-learning round roughly 20 times faster than feature-diversity selection on Cityscapes in the paper's measurements.
  • The spatial term is compatible with uncertainty and feature diversity and can be added to existing selection objectives without extra training data.
  • Because the objective is a max-min diversification problem, the same greedy solver applies to any region representation, including superpixels or irregular patches.
  • The paper's fixed-budget comparison suggests that a simpler explicit selection rule can beat a learned reinforcement-learning policy while using less annotation budget.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If spatial distance is a proxy for semantic diversity, the method should transfer to settings where categories correlate with location, such as street scenes, but lose its edge on datasets with objects uniformly repeated everywhere; that is an empirical prediction the paper does not test.
  • One could make the spatial penalty adaptive by shrinking $\tau$ as the labeling budget grows, which the paper's own sensitivity experiments suggest may help late-stage selection.
  • The same max-min objective could be applied to whole-image or video-frame selection, where 'spatial' becomes temporal or geometric distance, a direct extension of the paper's argument.
  • Because the method needs no feature extractor and no extra labeled data, it is a natural plug-in for settings with domain shift, where a pretrained feature extractor would be unreliable.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes Entropy+Spatial, a region-based active learning method for semantic segmentation that augments an entropy uncertainty term with a coordinate-based spatial diversity term inside a max-min batch-selection objective solved greedily. The spatial distance is a piecewise constant function (Eq. 6) that penalizes nearby same-image regions and assigns equal distance to far same-image and cross-image regions. Experiments on Cityscapes and PASCAL VOC 2012 compare against Random, Entropy, Core-Set, Entropy+Feature, and combined variants, and include ablations of the distance form, tau, a, b, c, region size, objective form, feature dimension, runtime, and image coverage. The method reports about 95-96% of fully supervised mIoU with 8.4%/5.9% labeled pixels and a fixed-budget improvement over RAL.

Significance. If the empirical results hold, the method is practically valuable: it replaces feature-based diversity with a cheap coordinate-only penalty, is compatible with uncertainty sampling, and scales to about 10^5 regions. The paper has notable strengths: controlled comparisons with 3-run means and standard deviations, a broad ablation suite, a computational cost analysis, and concrete performance claims. The central relative improvement over Entropy and Entropy+Feature appears robust. However, the formal statement of the optimization objective is flawed, and the mechanism behind the spatial term is not fully isolated from image-level coverage, so the conceptual contribution needs additional support before the paper can be accepted as is.

major comments (3)
  1. [Section III-A, Eq. (3)] The objective as displayed is not well posed. Written literally, min_{x_k in B_t, x_i,x_j in L_t union B_t} [lambda_u u(x_k)+d(x_i,x_j)] allows x_i=x_j, making d=0, and the pairwise term is independent of x_k; the minimum can collapse to the entropy-only term. The greedy criterion in Eq. (4) is coherent, but it is not the reformulation claimed in the text. Please restate Eq. (3) in the form max_{B_t} min_{x_k in B_t} [lambda_u u(x_k) + min_{x_i in L_t union B_t, x_i != x_k} d(x_i,x_k)] (or an equivalent max-min facility-dispersion objective) and adjust the set-function definitions in Eq. (2) accordingly.
  2. [Section III-B, Eq. (6), and Section IV-F] With the default a=1, b=2, c=2, Eq. (6) assigns the same distance (2) to any candidate in an image without a selected region and to any same-image candidate farther than tau from the selected set; only same-image candidates within tau receive the smaller penalty. Consequently, the objective contains a strong image-coverage component, and the claimed 'local spatial diversity' mechanism is not isolated. The ablation in Section V-C showing that varying c/b has little effect argues against image coverage being the sole driver, but the paper never compares Entropy+Spatial against an image-coverage-only baseline or a local-penalty-only baseline, nor does it report a direct relationship between d_s and label diversity. Please add such ablations (for example, entropy plus a constant per-image penalty, and entropy plus a same-image-only local penalty) and/or measure the label diversity of the selected sets.
  3. [Section IV-D, Table I] The fixed-budget comparison with RAL is imported from [7] rather than rerun under the paper's training protocol, and RAL's budget includes 350 labeled images used to train its policy. The text's 'for fairness' claim is not fully supported: differences in training iterations, augmentation, and pre-training may affect the comparison. The abstract's claim of outperforming all state-of-the-art region-based AL methods rests partly on this table. Either rerun RAL under the same protocol or qualify the claim to the controlled experiments in Fig. 2.
minor comments (5)
  1. [Abstract/Conclusion vs. Section IV-D] The percentage targets are inconsistent: the abstract and conclusion say 95% with 8.4%/5.9% labeled pixels, while Section IV-D says 'more than 96%' at 8.4% on Cityscapes. Please align the claims.
  2. [Section III-B, Eq. (5)] Eq. (5) defines d_s using coordinates loc_i but does not state how coordinates are normalized or how distances between regions in different images are computed. Please clarify, since the linear variant in Table II is compared with the piecewise variant.
  3. [Algorithm 1] Algorithm 1 uses 'while t <= T' with T as the maximum number of batches. If batches are indexed 0 through 5, the loop should likely be t < T, or T should be defined accordingly.
  4. [Table IV] Table IV reports the number of annotated images but does not explicitly state that the numbers are for Cityscapes; please state the dataset and, if available, provide the PASCAL VOC counterpart or explain its omission.
  5. [Section IV-B] There is a typo in the fully supervised training details: '0.007 for VOC 2012 and .' should be completed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's central claim is an empirical benchmark result, not a quantity forced by its own definitions or fit.

full rationale

The paper contains no derivation chain that reduces to its inputs. The central claim, that adding spatial diversity to entropy-based region active learning improves mIoU and reaches 95% of fully supervised performance with 8.4% and 5.9% labeled pixels, is a set of measured benchmark outcomes rather than a quantity predicted from fitted constants. Equation (6) defines a piecewise spatial distance with fixed constants a=1, b=2, c=2, and tau=N; this is an input heuristic. The greedy selection in Eq. (4) is an optimization procedure, and the resulting selected regions are then used to retrain the network and measure mIoU on held-out validation splits. No step fits a parameter to a subset of the reported mIoU curve and then predicts that same curve. The external references used in the unified framework, such as Core-Set and USDM, are cited to instantiate existing objectives, not to justify the paper's own results by a self-citation chain. The skeptical observation that Eq. (6) assigns equal distance to same-image far pairs and cross-image pairs, so that the objective partly maximizes image coverage, is a mechanism-interpretation concern rather than circularity: the empirical improvement could still hold even if the stated 'local diversity' rationale is incomplete. Likewise, hyperparameter choices and ablations in Section V are evaluated on the same benchmark datasets, which is a selection-bias caveat, but the reported outcome is a measured result rather than a definitional or fitted equality. Thus no circular step is exhibited in the manuscript.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The method does not introduce new physical or conceptual entities. It introduces a hand-crafted distance function and objective weights, which are captured as free parameters. The core assumptions are the spatial-label diversity correlation and the reliability of entropy as an uncertainty signal, both standard but untested independently.

free parameters (4)
  • tau (τ) = N (region size); 128 for Cityscapes, 32 for PASCAL VOC
    Threshold defining which regions are considered close in Eq. (6). Chosen as the region size; Fig. 5 shows the optimal value varies with dataset and budget.
  • a, b, c = 1, 2, 2
    Distance values in the piecewise function Eq. (6). Only ratios b/a and c/b matter after normalization; set to b/a=2, c/b=1. Fig. 6 shows modest sensitivity, and the values are hand-picked, not derived.
  • lambda_u, lambda_f, lambda_s = 1, 0 or 1, 1 depending on variant
    Weights for uncertainty, feature diversity, and spatial diversity in Eq. (3). The final Entropy+Spatial method uses (1, 0, 1). No explicit tuning is reported; terms are normalized to [0,1] before summation.
  • Region size N = 128 (Cityscapes), 32 (PASCAL VOC)
    The grid size for dividing images into regions. It affects both the spatial distance scale (tau=N) and the granularity of annotation. Fig. 7 shows that smaller N helps the proposed method, so the chosen values favor the method.
assumptions (4)
  • standard math Greedy selection via potential function Eq. (4) approximately solves the max-min objective Eq. (3).
    The paper relies on the greedy algorithm from [24] for maximin diversification; no approximation guarantee or submodularity analysis is given for the specific piecewise spatial distance.
  • domain assumption Spatially distant regions tend to cover different semantic categories.
    Stated in Section III-B: enforcing spatial diversity 'aims to enforce selected regions to be diverse in the image domain therefore there is a high chance of covering diverse categories.' The entire method depends on this correlation.
  • domain assumption Mean pixel entropy from the current segmentation network is a good uncertainty measure for region selection.
    Section III-A defines u(x) as the average of pixel entropies; this is the standard AL assumption that uncertain regions are informative, borrowed without dedicated justification.
  • standard math The piecewise distance function is a metric for the chosen parameters with c >= b >= a > 0 and b <= 2a.
    Appendix A proves the triangle inequality under these constraints; the selected values a=1, b=2, c=2 satisfy the condition, enabling its use in the max-min objective.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Exploring Spatial Diversity for Region-based Active Learning." pith.science (2026). https://pith.science/paper/TBZUJ5OT

@misc{pith2026250717367,
  author       = {Pith},
  title        = {Pith review of: Exploring Spatial Diversity for Region-based Active Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TBZUJ5OT}},
  note         = {Machine review of arXiv:2507.17367}
}
abstract

State-of-the-art methods for semantic segmentation are based on deep neural networks trained on large-scale labeled datasets. Acquiring such datasets would incur large annotation costs, especially for dense pixel-level prediction tasks like semantic segmentation. We consider region-based active learning as a strategy to reduce annotation costs while maintaining high performance. In this setting, batches of informative image regions instead of entire images are selected for labeling. Importantly, we propose that enforcing local spatial diversity is beneficial for active learning in this case, and to incorporate spatial diversity along with the traditional active selection criterion, e.g., data sample uncertainty, in a unified optimization framework for region-based active learning. We apply this framework to the Cityscapes and PASCAL VOC datasets and demonstrate that the inclusion of spatial diversity effectively improves the performance of uncertainty-based and feature diversity-based active learning methods. Our framework achieves $95\%$ performance of fully supervised methods with only $5-9\%$ of the labeled pixels, outperforming all state-of-the-art region-based active learning methods for semantic segmentation.

Figures

Figures reproduced from arXiv: 2507.17367 by the authors.

Figure 1
Figure 1. We propose a new objective, i.e., spatial diversity, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Segmentation performance vs. amount of annotated data on Cityscapes (left) and PASCAL VOC 2012 (right). We report [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. As neighboring regions usually have similar entropy [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (6 more)
Figure 3
Figure 3. Figure 3: Visualization of regions selected by different methods at the second batch (total budget: 2k regions). Note how the [PITH_FULL_IMAGE:figures/full_fig_p007_3.png]
Figure 4
Figure 4. Figure 4: Visualization of selected regions of Entropy+Spatial with different annotation budgets. Note how our method allows [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Effect of τ on performance for Cityscapes (left) and PASCAL VOC 2012 (right). C. Effect of a, b and c on Performance For the piece-wise distance function defined in Eq. (6), parameters a, b and c specify the value for each piece. As each [PITH_FULL_IMAGE:figures/full_…
Figure 7
Figure 7. Figure 7: Effect of N on performance for Cityscapes (left) and PASCAL VOC 2012 (right). TABLE III: Comparing Max-Min and Max-Sum as objective with Entropy+Spatial on Cityscapes and PASCAL VOC 2012. Both the mean and standard deviation of 3 runs are reported. Budget 2k 4k 8k 16k …
Figure 6
Figure 6. Figure 6: Sensitivity of a, b, c on performance for Cityscapes (left) and PASCAL VOC 2012 (right). D. Effect of N on Performance To study the effect of N (region size), we keep all other parameters the same, i.e., τ = N, a = 1, b = c = 2. Also, the amount of pixels selected at e…
Figure 8
Figure 8. Figure 8: Effect of feature dimension on Entropy+Feature for [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 28 canonical work pages

  1. [7]

    Rein- forced Active Learning for Image Segmentation,

    A. Casanova, P. Pinheiro, N. Rostamzadeh, and P. Christopher, “Rein- forced Active Learning for Image Segmentation,” in ICLR, 2020

    work page 2020

  2. [1]

    Settles, Active Learning, ser

    B. Settles, Active Learning, ser. Synthesis Lectures on Artificial Intelli- gence and Machine Learning. Morgan & Claypool, 2012

    work page 2012

  3. [2]

    Multi-class active learning by uncertainty sampling with diversity maximization,

    Y . Yang, Z. Ma, F. Nie, X. Chang, and A. G. Hauptmann, “Multi-class active learning by uncertainty sampling with diversity maximization,” International Journal of Computer Vision , 2015

    work page 2015

  4. [3]

    Deep bayesian active learning with image data,

    Y . Gal, R. Islam, and Z. Ghahramani, “Deep bayesian active learning with image data,” in ICML, 2017

    work page 2017

  5. [4]

    Active learning for convolutional neural networks: A core-set approach,

    O. Sener and S. Savarese, “Active learning for convolutional neural networks: A core-set approach,” in ICLR, 2018

    work page 2018

  6. [5]

    The cityscapes dataset for semantic urban scene understanding,

    M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Benen- son, U. Franke, S. Roth, and B. Schiele, “The cityscapes dataset for semantic urban scene understanding,” in CVPR, 2016

    work page 2016

  7. [6]

    CEREALS - cost-effective region-based active learning for semantic segmentation,

    R. Mackowiak, P. Lenz, O. Ghori, F. Diego, O. Lange, and C. Rother, “CEREALS - cost-effective region-based active learning for semantic segmentation,” in BMVC, 2018

    work page 2018

  8. [8]

    A sequential algorithm for training text classifiers,

    D. D. Lewis and W. A. Gale, “A sequential algorithm for training text classifiers,” in Proceedings of the 17th Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval. Dublin, Ireland, 3-6 July 1994 (Special Issue of the SIGIR Forum), 1994

    work page 1994

Show all 33 references
  1. [9]

    Multi-class active learning for image classification,

    A. J. Joshi, F. Porikli, and N. Papanikolopoulos, “Multi-class active learning for image classification,” in CVPR, 2009

    2009

  2. [10]

    Bayesian active learning for classification and preference learning,

    N. Houlsby, F. Husz ´ar, Z. Ghahramani, and M. Lengyel, “Bayesian active learning for classification and preference learning,” arXiv preprint arXiv:1112.5745, 2011

    2011 arXiv

  3. [11]

    Active learning using pre- clustering,

    H. T. Nguyen and A. W. M. Smeulders, “Active learning using pre- clustering,” in ICML, 2004

    2004

  4. [12]

    Querying discriminative and representative samples for batch mode active learning,

    Z. Wang and J. Ye, “Querying discriminative and representative samples for batch mode active learning,” in ACM SIGKDD, 2013

    2013

  5. [13]

    Active learning for semantic segmentation with expected change,

    A. Vezhnevets, J. M. Buhmann, and V . Ferrari, “Active learning for semantic segmentation with expected change,” in CVPR, 2012

    2012

  6. [14]

    The power of ensembles for active learning in image classification,

    W. H. Beluch, T. Genewein, A. N ¨urnberger, and J. M. K ¨ohler, “The power of ensembles for active learning in image classification,” inCVPR, 2018

    2018

  7. [15]

    Learning loss for active learning

    D. Yoo and I. S. Kweon, “Learning loss for active learning.” in CVPR, 2019, pp. 93–102

    2019

  8. [16]

    Fine-tuning convolutional neural networks for biomedical image analysis: Actively and incrementally,

    Z. Zhou, J. Y . Shin, L. Zhang, S. R. Gurudu, M. B. Gotway, and J. Liang, “Fine-tuning convolutional neural networks for biomedical image analysis: Actively and incrementally,” in CVPR, 2017

    2017

  9. [17]

    Active image segmentation propagation,

    S. Dutt Jain and K. Grauman, “Active image segmentation propagation,” in CVPR, 2016

    2016

  10. [18]

    Suggestive annotation: A deep active learning framework for biomedical image segmentation,

    L. Yang, Y . Zhang, J. Chen, S. Zhang, and D. Z. Chen, “Suggestive annotation: A deep active learning framework for biomedical image segmentation,” in Medical Image Computing and Computer Assisted Intervention - MICCAI 2017 - 20th International Conference, Quebec City, QC, Can...

    2017

  11. [19]

    Efficient active learning for image classification and segmentation using a sample selection and conditional generative adversarial network,

    D. Mahapatra, B. Bozorgtabar, J.-P. Thiran, and M. Reyes, “Efficient active learning for image classification and segmentation using a sample selection and conditional generative adversarial network,” in Interna- tional Conference on Medical Image Computing and Computer-Assist...

    2018

  12. [20]

    Variational adversarial active learning,

    S. Sinha, S. Ebrahimi, and T. Darrell, “Variational adversarial active learning,” in Proceedings of the IEEE International Conference on Computer Vision, 2019, pp. 5972–5981

    2019

  13. [21]

    Region-based active learning for efficient labeling in se- mantic segmentation,

    T. Kasarla, G. Nagendar, G. M. Hegde, V . Balasubramanian, and C. Jawahar, “Region-based active learning for efficient labeling in se- mantic segmentation,” in 2019 IEEE Winter Conference on Applications of Computer Vision (WACV) . IEEE, 2019, pp. 1109–1117

    2019

  14. [22]

    Viewal: Active learning with viewpoint entropy for semantic segmentation,

    Y . Siddiqui, J. Valentin, and M. Nießner, “Viewal: Active learning with viewpoint entropy for semantic segmentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 9433–9443

    2020

  15. [23]

    Heuristic and special case algorithms for dispersion problems,

    S. S. Ravi, D. J. Rosenkrantz, and G. K. Tayi, “Heuristic and special case algorithms for dispersion problems,” Operations Research, 1994

    1994

  16. [24]

    An axiomatic approach for result diversi- fication,

    S. Gollapudi and A. Sharma, “An axiomatic approach for result diversi- fication,” in Proceedings of the 18th international conference on World wide web, 2009

    2009

  17. [25]

    A survey of deep active learning,

    P. Ren, Y . Xiao, X. Chang, P.-Y . Huang, Z. Li, X. Chen, and X. Wang, “A survey of deep active learning,” arXiv preprint arXiv:2009.00236 , 2020

    2009 arXiv

  18. [26]

    The PASCAL Visual Object Classes Challenge 2012 (VOC2012) Results,

    M. Everingham, L. Van Gool, C. K. I. Williams, J. Winn, and A. Zisserman, “The PASCAL Visual Object Classes Challenge 2012 (VOC2012) Results,” http://www.pascal- network.org/challenges/VOC/voc2012/workshop/index.html

    2012

  19. [27]

    Auto-deeplab: Hierarchical neural architecture search for semantic image segmentation,

    C. Liu, L.-C. Chen, F. Schroff, H. Adam, W. Hua, A. L. Yuille, and L. Fei-Fei, “Auto-deeplab: Hierarchical neural architecture search for semantic image segmentation,” in CVPR, 2019

    2019

  20. [28]

    Encoder- decoder with atrous separable convolution for semantic image segmen- tation,

    L.-C. Chen, Y . Zhu, G. Papandreou, F. Schroff, and H. Adam, “Encoder- decoder with atrous separable convolution for semantic image segmen- tation,” in ECCV, 2018

    2018

  21. [29]

    Xception: Deep learning with depthwise separable convo- lutions,

    F. Chollet, “Xception: Deep learning with depthwise separable convo- lutions,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017, Honolulu, HI, USA, July 21-26, 2017 , 2017, pp. 1800–1807

    2017

  22. [30]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in 2009 IEEE conference on computer vision and pattern recognition . Ieee, 2009, pp. 248–255

    2009

  23. [31]

    Feature pyramid networks for object detection,

    T.-Y . Lin, P. Doll´ar, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature pyramid networks for object detection,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2017, pp. 2117–2125

    2017

  24. [32]

    Playing for data: Ground truth from computer games,

    S. R. Richter, V . Vineet, S. Roth, and V . Koltun, “Playing for data: Ground truth from computer games,” in European conference on com- puter vision. Springer, 2016, pp. 102–118

    2016

  25. [33]

    Max-sum diversification, monotone submodular functions and dynamic updates,

    A. Borodin, H. C. Lee, and Y . Ye, “Max-sum diversification, monotone submodular functions and dynamic updates,” in Proceedings of the 31st ACM SIGMOD-SIGACT-SIGAI symposium on Principles of Database Systems, 2012, pp. 155–166. Lile Cai received her Bachelor of Engineering fro...

    2012

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.