Pith. sign in

REVIEW 4 major objections 7 minor 61 references

B-FPGM: Lightweight Face Detection via Bayesian-Optimized Soft FPGM Pruning

T0 review · 4 major / 7 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read This paper claims that Bayesian optimization of per-layer-group pruning rates, combined with geometric-median soft pruning, yields a better size-versus-accuracy trade-off for tiny face detectors than applying a single uniform pruning rate…

desk verdict Useful BO-per-group pruning result for tiny face detectors, but the 'consistently superior' and 'first BO structured pruning' claims both overreach; worth a serious referee after fixes. read the letter →

arxiv 2501.16917 v1 pith:57EWKAUU submitted 2025-01-28 cs.CV

Nikolaos Kaparinos , Vasileios Mezaris This is my paper · ORCID
classification cs.CV
keywords facedetectionnetworkpruningBayesianoptimizationfiltergeometricmediansoftlightweightmodelsWIDER
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 tries to show that the per-layer pruning-rate decisions in structured pruning of face detectors should not be uniform, and that Bayesian optimization can discover better non-uniform rates automatically. The authors combine geometric-median filter pruning (FPGM) with soft filter pruning (SFP) and a Bayesian search over layer-group pruning rates, then apply the result to two of the smallest competitive face detectors, EResFD and EXTD, evaluating on WIDER FACE. Across target sparsities from 10% to 60%, the Bayesian-optimized versions consistently keep more detection accuracy (mAP) than the same pipeline pruned uniformly at the same or even smaller model size. The payoff is that edge-device face detectors can be compressed substantially, up to roughly 60% of parameters, with a much gentler accuracy decline than uniform pruning delivers, and the work is the first to apply Bayesian optimization to structured pruning for face detection.

What carries the argument

The machinery is a three-stage pipeline. First, the network is partitioned into $N$ layer groups, with a default of 6, to keep the search space small. Second, Bayesian optimization with an upper-confidence-bound acquisition function searches for the vector of per-group pruning rates $\phi^*$, guided by an objective that combines the validation loss after a single-epoch soft-pruned training run with a penalty for missing the target sparsity. Third, the optimized rates are applied through the FPGM/SFP procedure: filters closest to the geometric median of each layer's filter set are iteratively soft-pruned, meaning zeroed but still updated, during retraining, then hard-pruned and fine-tuned. The geometric median acts as a proxy for redundant filters, SFP lets the network recover from pruning during training, and Bayesian optimization removes the need to hand-engineer layer-wise rates.

What would settle it

Take a fixed target sparsity, for example 20%, and generate many random pruning-rate vectors; for each, record the one-epoch validation-loss proxy and then run the complete soft-prune, retrain, hard-prune, and fine-tune pipeline to get final WIDER FACE mAP. If the rank correlation between the proxy and the final mAP is weak, the optimization objective is not carrying the argument. A direct version is to find a random vector that scores worse on the proxy but, after full training, beats the B-FPGM rates at matched sparsity.

Watch

Extended reading notes

Core claim

The central claim is that replacing the uniform pruning vector $\phi=[T,\ldots,T]$ with a per-group vector $\phi^*$ found by Bayesian optimization changes the size–performance trade-off of FPGM-based soft pruning in a way that dominates uniform pruning for lightweight face detectors. On EResFD at target sparsities 10–60%, B-FPGM achieves mAP on WIDER FACE that is close to the original model at sparsities up to 50%, while uniform FPGM collapses at 50% and 60%, with Hard-subset mAP falling from 0.7757 to 0.5254 and 0.2936, versus B-FPGM's 0.6993 and 0.6051 at slightly higher actual sparsity. The authors attribute the advantage to the optimizer concentrating pruning in later layer groups, Groups 3–6 of EResFD, where redundancy is higher, while sparing early layers. A smaller ablation on EXTD and a grouping ablation support the claim that the method transfers across architectures.

Load-bearing premise

The entire rate search stands on the assumption that the validation loss measured after one short training epoch on a trial-pruned network reliably predicts how accurate the fully pruned and fine-tuned model will end up being.

Editorial extensions

If this is right

  • Tiny face detectors can be compressed to around half their parameters while keeping practical accuracy, making them more deployable on phones, drones, and other edge devices.
  • The improvement is largest at high sparsities of 50–60%, where uniform pruning fails but B-FPGM retains usable detection, so the method extends the usable compression range.
  • Because the optimizer automatically chooses which layers to prune, engineering effort shifts from manual rate tuning to defining layer groups and a reliable validation-loss proxy.
  • The authors show transfer to a second architecture, EXTD, suggesting the method does not depend on EResFD-specific details.

Reading between the lines

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

  • The one-epoch surrogate may favor conservative pruning in early layers because loss after one epoch is dominated by layers that are hard to recover, so a multi-epoch or learned surrogate could discover even better allocations.
  • The method's reliance on hand-defined layer groups leaves on the table some of the benefit of full per-layer optimization, and automatic grouping by sensitivity could push the trade-off further.
  • Because the Bayesian optimizer only sees a single epoch, rates that look good early may be suboptimal after the full 200-epoch soft-pruning schedule, and ablating the number of surrogate epochs would test this.
  • Applying the same optimization to detection heads or to latency-aware objectives such as FLOPs or inference time, rather than parameter count alone, could yield different pruning patterns.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes B-FPGM, a structured pruning pipeline for lightweight face detectors that combines FPGM filter pruning, Soft Filter Pruning (SFP), and Bayesian optimization. Instead of applying a uniform pruning rate to all layers, the method groups layers (N=6 by default), uses Bayesian optimization to select per-group pruning rates for a target sparsity T, and then applies the optimized rates within the SFP/FPGM training scheme of Gkrispanis et al. Experiments on the WIDER FACE benchmark compare B-FPGM with uniform FPGM pruning on the EResFD detector across targets from 10% to 60%, plus a smaller EXTD ablation and grouping ablations. The paper claims that B-FPGM consistently achieves a superior size-to-performance trade-off, with code and trained models publicly available.

Significance. If the empirical claims hold, the paper offers a practical contribution: automated per-group pruning-rate selection for small face detectors, with source code and trained models, a two-architecture demonstration, and a five-seed stability check at T=20%. The method is simple and reproducible in principle. However, the evidence as presented is not yet sufficient to support the central claim of consistent superiority: the comparison is made at mismatched actual sparsity levels, the low-sparsity regime contradicts the claim, and only a single pruning baseline is used. The core idea is plausible and the high-sparsity gains are large, but those gains need to be separated from the confound of larger effective pruning budgets.

major comments (4)
  1. [Table 3, T=10% row] The stated central claim that B-FPGM 'consistently' achieves a superior size-to-performance trade-off is directly contradicted by the T=10% row. At this target, uniform FPGM reaches mAP 0.8728/0.8582/0.7757 on Easy/Medium/Hard at 5.25% actual sparsity, while B-FPGM reaches only 0.8622/0.8506/0.7636 at 10.24% actual sparsity. B-FPGM is therefore both smaller and worse on every subset at low sparsity. The claim must be restricted to moderate-to-high pruning rates, or a low-sparsity regime must be added where B-FPGM is not worse.
  2. [Tables 3 and 4, Fig. 3] The comparison is confounded by systematically unmatched actual sparsity. For every target T, B-FPGM's actual sparsity is higher than uniform FPGM's (e.g., T=20: 22.27% vs 16.84%; T=60: 59.87% vs 54.05%). Fig. 3 plots nominal target sparsity, not actual sparsity, so the visible gains at high T may partly reflect that B-FPGM is being compared at a larger effective pruning budget. The authors should provide a size-matched comparison, for example by plotting mAP against actual sparsity on the horizontal axis or by running uniform FPGM at the same actual sparsity as each B-FPGM model, and should report the actual sparsity values on the plots.
  3. [Section 3.3, Eq. (1)] The entire Bayesian optimization step rests on the assumption that the validation loss after a trial soft-pruning and one epoch of training is a reliable proxy for the final mAP after the full soft-pruning/retraining and hard-pruning/fine-tuning pipeline. The paper provides no evidence for this correlation, such as a scatter plot of proxy values versus final performance for a sample of pruning-rate vectors. If this one-epoch proxy is noisy or misleading, the optimized rates could be worse than uniform rates, which would invalidate the central comparison. To make the method trustworthy, the authors should validate the proxy empirically or provide a theoretical/empirical argument for why one epoch suffices.
  4. [Section 4.2 and Fig. 4] The Abstract and Section 5 claim superiority over 'existing approaches' and 'SoA' in balancing size and performance, but the pruning comparison in Tables 3 and 4 is only against uniform FPGM pruning. Fig. 4 compares against other face detectors, not other pruning methods. To support the broader claim, the authors should either add at least one additional pruning baseline (e.g., L1-norm pruning, AMC-style automatic pruning, or a magnitude-based non-uniform baseline) or explicitly restrict the claim to 'superior to uniform FPGM at moderate-to-high sparsity on EResFD and EXTD.'
minor comments (7)
  1. [Algorithm 1] The pseudocode loop 'while i <= I' never increments i, so the algorithm as written would run forever; add an explicit 'i ← i + 1' at the end of the loop body.
  2. [Section 3.3] The acquisition function is called 'UPC' but is described as the Upper Confidence Bound; it should be 'UCB' to match standard terminology.
  3. [Table 4] The first row is labeled 'EXTD (orig.) [23]' but EXTD is reference [56], not [23]; the citation should be corrected.
  4. [Table 2 and Section 3.3] The notation for the bounds is unclear: the text says the bounds are '[0, T+ bound of f set]' and Table 2 lists 'T + 0.04' and 'bound of f set 0.2' without a clear formula. Please define the bound explicitly, e.g., upper bound = T + offset, and state the offset value in a consistent notation.
  5. [Eq. (1)] The condition 'T − T + ≤ S(θϕi ) ≤ T + T +' is visually ambiguous because 'T+' is used both as a threshold symbol and an addition operator; use an explicit threshold variable, such as τ, to avoid confusion.
  6. [Tables 3, 4, and Fig. 3] The main results tables and figures do not report error bars or multiple-seed variability, even though Table 6 shows that B-FPGM has moderate variance at T=20% (std up to 0.0066 on Hard). Reporting the five-seed mean and standard deviation for all main results would strengthen the comparison.
  7. [References [59] and [60]] References [59] and [60] appear to be the same paper, but they are listed as separate entries and cited differently (Eagle Eye [59] in the text vs [60] in Fig. 4); this should be cleaned up.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the pruning-rate search is validated on held-out data and the final comparison is external; the unmatched actual sparsity is an evaluation-validity issue, not circularity.

full rationale

The paper's central claim is an empirical comparison of pruning-rate allocation strategies. The Bayesian optimizer (Eq. 1) selects per-group pruning rates using a held-out 20% split of the training set (Sec. 4.1) and an objective that combines the one-epoch validation loss with a sparsity penalty; the final models are then retrained and hard-pruned, and evaluated on the separate WIDER FACE validation set. No fitted parameter is relabeled as a prediction, and the final evaluation metric (mAP@0.5) is external to the optimization objective. The only self-citation is to the authors' prior FPGM pruning pipeline [12], which is used as the uniform-pruning baseline and as the retraining procedure; this is a baseline, not an imported uniqueness theorem or an unverified premise, so it is not load-bearing circularity. The comparison at a fixed nominal target T is weakened by the fact that B-FPGM's actual sparsity is systematically higher than uniform FPGM's at the same T (e.g., 10.24% vs 5.25% at T=10%), and B-FPGM is worse at that row; this is a correctness/validity concern about the trade-off claim, not a circularity of the derivation.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a small number of hand-chosen design elements: the layer grouping, the BO hyperparameters, and the one-epoch validation-loss proxy that guides the search. The FPGM and SFP priors are inherited from cited work. No new entities are introduced.

free parameters (3)
  • Layer grouping (N=6 assignment of layers to groups) = N=6 for EResFD and EXTD (details in Fig. 2)
    Chosen ad hoc by inspecting architecture diagrams (Section 3.2); the ablation with N=4 and N=9 shows N=6 works best, but the grouping itself is hand-designed and not learned.
  • BO hyperparameters (I, i0, lambda, penalty value, bound offset) = I=1000, i0=60, lambda=5, penalty=100, T+ deviation threshold and bound offset in Table 2
    These are set manually (Table 2) and not tuned or justified; the upper bound of each pruning rate appears binding at T=60% where rates hit the 80% cap.
  • Target sparsity deviation threshold T+ = 0.04 (from Table 2)
    Used in the acceptability condition in Eq. 1; no sensitivity analysis is given.
assumptions (4)
  • domain assumption Filters closest to the geometric median are the most redundant and can be pruned without significant accuracy loss (FPGM criterion).
    Adopted from [19] and [12]; used in Section 3.4 to select filters for pruning in each layer.
  • domain assumption Soft Filter Pruning with periodic re-pruning and retraining produces a well-performing final network (SFP paradigm).
    Adopted from [17] and [12]; the training loop in Algorithm 2 follows this paradigm.
  • ad hoc to paper The validation loss after one epoch of training on a trial-pruned network is a good proxy for the final performance after full retraining.
    This is the objective function f_theta in Eq. 1 (Section 3.3). If this proxy is poor, the Bayesian optimization may select suboptimal pruning rates. No validation of the proxy is provided.
  • ad hoc to paper The hand-designed layer grouping preserves enough flexibility to capture the different sensitivities of layers.
    The grouping into N=6 groups is made ad hoc (Section 3.2). The method's success depends on this grouping being a useful search space.

how reviews work

0 comments
Cite this review

Pith. "Pith review of B-FPGM: Lightweight Face Detection via Bayesian-Optimized Soft FPGM Pruning." pith.science (2026). https://pith.science/paper/57EWKAUU

@misc{pith2026250116917,
  author       = {Pith},
  title        = {Pith review of: B-FPGM: Lightweight Face Detection via Bayesian-Optimized Soft FPGM Pruning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/57EWKAUU}},
  note         = {Machine review of arXiv:2501.16917}
}
read the original abstract

Face detection is a computer vision application that increasingly demands lightweight models to facilitate deployment on devices with limited computational resources. Neural network pruning is a promising technique that can effectively reduce network size without significantly affecting performance. In this work, we propose a novel face detection pruning pipeline that leverages Filter Pruning via Geometric Median (FPGM) pruning, Soft Filter Pruning (SFP) and Bayesian optimization in order to achieve a superior trade-off between size and performance compared to existing approaches. FPGM pruning is a structured pruning technique that allows pruning the least significant filters in each layer, while SFP iteratively prunes the filters and allows them to be updated in any subsequent training step. Bayesian optimization is employed in order to optimize the pruning rates of each layer, rather than relying on engineering expertise to determine the optimal pruning rates for each layer. In our experiments across all three subsets of the WIDER FACE dataset, our proposed approach B-FPGM consistently outperforms existing ones in balancing model size and performance. All our experiments were applied to EResFD, the currently smallest (in number of parameters) well-performing face detector of the literature; a small ablation study with a second small face detector, EXTD, is also reported. The source code and trained pruned face detection models can be found at: https://github.com/IDTITI/B-FPGM.

Figures

Figures reproduced from arXiv: 2501.16917 by the authors.

Figure 1
Figure 1. Overview of our proposed pruning and training pipeline. The diagram on the left illustrates our complete methodology, while the [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Layer grouping of (a) the EResFD model, (b) the EXTD [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Comparative results (mAP) on the (a) Easy, (b) Medium, (c) Hard WIDER FACE subsets. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Model size and mAP across different face detector models on the Hard subset of WIDER FACE. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Examples of face detection results using the original [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 57 canonical work pages

  1. [1]

    FDLite: A Single Stage Lightweight Face Detector Network

    Yogesh Aggarwal and Prithwijit Guha. FDLite: A sin- gle stage lightweight face detector network. arXiv preprint arXiv:2406.19107, 2024

    work page Pith review arXiv 2024

  2. [2]

    Struc- tured pruning of deep convolutional neural networks

    Sajid Anwar, Kyuyeon Hwang, and Wonyong Sung. Struc- tured pruning of deep convolutional neural networks. ACM Journal on Emerging Technologies in Computing Systems (JETC), 13(3):1–18, 2017

    work page 2017

  3. [3]

    Efficient neural net- work pruning using model-based reinforcement learning

    Blanka Bencsik and M ´arton Szemenyei. Efficient neural net- work pruning using model-based reinforcement learning. In 2022 Int. Symp. on Measurement and Control in Robotics (ISMCR), pages 1–8. IEEE, 2022

    work page 2022

  4. [4]

    The exploration-exploitation dilemma: a multidisci- plinary framework

    Oded Berger-Tal, Jonathan Nathan, Ehud Meron, and David Saltz. The exploration-exploitation dilemma: a multidisci- plinary framework. PloS one, 9(4):e95693, 2014

    work page 2014

  5. [5]

    Learning compact representations of neural networks using discrim- inative masking (DAM)

    Jie Bu, Arka Daw, M Maruf, and Anuj Karpatne. Learning compact representations of neural networks using discrim- inative masking (DAM). Advances in Neural Information Processing Systems, 34:3491–3503, 2021

    work page 2021

  6. [6]

    A survey on deep neural network pruning: Taxonomy, compar- ison, analysis, and recommendations

    Hongrong Cheng, Miao Zhang, and Javen Qinfeng Shi. A survey on deep neural network pruning: Taxonomy, compar- ison, analysis, and recommendations. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    work page 2024

  7. [7]

    Basic enhancement strategies when using bayesian optimization for hyperparam- eter tuning of deep neural networks

    Hyunghun Cho, Yongjin Kim, Eunjung Lee, Daeyoung Choi, Yongjae Lee, and Wonjong Rhee. Basic enhancement strategies when using bayesian optimization for hyperparam- eter tuning of deep neural networks. IEEE Access, 8:52588– 52608, 2020

    work page 2020

  8. [8]

    Retinaface: Single-shot multi-level face localisation in the wild

    Jiankang Deng, Jia Guo, Evangelos Ververas, Irene Kotsia, and Stefanos Zafeiriou. Retinaface: Single-shot multi-level face localisation in the wild. In Proc. CVPR, pages 5203– 5212, 2020

    work page 2020

Show all 61 references
  1. [9]

    Real-time face detection and tracking on mobile phones for criminal detection

    Lamiaa A Elrefaei, Alaa Alharthi, Huda Alamoudi, Shatha Almutairi, and Fatima Al-rammah. Real-time face detection and tracking on mobile phones for criminal detection. In 2017 2nd Int. Conf. on Anti-Cyber Crimes (ICACC) , pages 75–80. IEEE, 2017

    2017

  2. [10]

    Depgraph: Towards any structural pruning

    Gongfan Fang, Xinyin Ma, Mingli Song, Michael Bi Mi, and Xinchao Wang. Depgraph: Towards any structural pruning. In Proc. CVPR, pages 16091–16101, 2023

    2023

  3. [11]

    A tutorial on bayesian optimization

    Peter I Frazier. A tutorial on bayesian optimization. arXiv preprint arXiv:1807.02811, 2018

    2018 arXiv

  4. [12]

    Filter-pruning of lightweight face detectors using a geometric median criterion

    Konstantinos Gkrispanis, Nikolaos Gkalelis, and Vasileios Mezaris. Filter-pruning of lightweight face detectors using a geometric median criterion. In Proc. WACV Workshops, pages 280–289, 2024

    2024

  5. [13]

    Fast and control- lable post-training sparsity: Learning optimal sparsity allo- cation with global constraint in minutes

    Ruihao Gong, Yang Yong, Zining Wang, Jinyang Guo, Xiuy- ing Wei, Yuqing Ma, and Xianglong Liu. Fast and control- lable post-training sparsity: Learning optimal sparsity allo- cation with global constraint in minutes. InProc. AAAI Conf. on Artificial Intelligence , volume 38, p...

    2024

  6. [14]

    Sample and computation redistribution for effi- cient face detection

    Jia Guo, Jiankang Deng, Alexandros Lattas, and Stefanos Zafeiriou. Sample and computation redistribution for effi- cient face detection. In Proc. ICLR, 2022

    2022

  7. [15]

    Learning to prune deep neural networks via reinforcement learning

    Manas Gupta, Siddharth Aravindan, Aleksandra Kalisz, Vi- jay Chandrasekhar, and Lin Jie. Learning to prune deep neural networks via reinforcement learning. arXiv preprint arXiv:2007.04756, 2020

    2007 arXiv

  8. [16]

    Cap: Context-aware pruning for semantic segmentation

    Wei He, Meiqing Wu, Mingfu Liang, and Siew-Kei Lam. Cap: Context-aware pruning for semantic segmentation. In Proc. WACV, pages 960–969, 2021

    2021

  9. [17]

    Soft filter pruning for accelerating deep convolutional neural networks

    Yang He, Guoliang Kang, Xuanyi Dong, Yanwei Fu, and Yi Yang. Soft filter pruning for accelerating deep convolutional neural networks. In IJCAI, pages 2234–2240, 2018

    2018

  10. [18]

    Amc: Automl for model compression and accel- eration on mobile devices

    Yihui He, Ji Lin, Zhijian Liu, Hanrui Wang, Li-Jia Li, and Song Han. Amc: Automl for model compression and accel- eration on mobile devices. In Proc. ECCV, pages 784–800, 2018

    2018

  11. [19]

    Filter pruning via Geometric Median for deep convolutional neural networks acceleration

    Yang He, Ping Liu, Ziwei Wang, Zhilan Hu, and Yi Yang. Filter pruning via Geometric Median for deep convolutional neural networks acceleration. In Proc. CVPR, 2019

    2019

  12. [20]

    LFFD: A light and fast face de- tector for edge devices

    Yonghao He, Dezhong Xu, Lifang Wu, Meng Jian, Shiming Xiang, and Chunhong Pan. LFFD: A light and fast face de- tector for edge devices. arXiv preprint arXiv:1904.10633 , 2019

    1904 arXiv

  13. [21]

    Mobilenets: Efficient convolu- tional neural networks for mobile vision applications

    Andrew G Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco An- dreetto, and Hartwig Adam. Mobilenets: Efficient convolu- tional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861, 2017

    2017 arXiv

  14. [22]

    Control of exploitation–exploration meta-parameter in reinforcement learning

    Shin Ishii, Wako Yoshida, and Junichiro Yoshimoto. Control of exploitation–exploration meta-parameter in reinforcement learning. Neural networks, 15(4-6):665–687, 2002

    2002

  15. [23]

    EResFD: Rediscovery of the effectiveness of standard convolution for lightweight face detection

    Joonhyun Jeong, Beomyoung Kim, Joonsang Yu, and YoungJoon Yoo. EResFD: Rediscovery of the effectiveness of standard convolution for lightweight face detection. In In Proc. WACV, pages 977–987, 2024

    2024

  16. [24]

    A review of yolo algorithm developments

    Peiyuan Jiang, Daji Ergu, Fangyao Liu, Ying Cai, and Bo Ma. A review of yolo algorithm developments. Procedia computer science, 199:1066–1073, 2022

    2022

  17. [25]

    Learning lightweight face detector with knowledge distillation

    Haibo Jin, Shifeng Zhang, Xiangyu Zhu, Yinhang Tang, Zhen Lei, and Stan Z Li. Learning lightweight face detector with knowledge distillation. In 2019 Int. Conf. on Biometrics (ICB), pages 1–7. IEEE, 2019

    2019

  18. [26]

    A review on genetic algorithm: past, present, and future

    Sourabh Katoch, Sumit Singh Chauhan, and Vijay Kumar. A review on genetic algorithm: past, present, and future. Mul- timedia tools and applications, 80:8091–8126, 2021

    2021

  19. [27]

    Face de- tection techniques: a review

    Ashu Kumar, Amandeep Kaur, and Munish Kumar. Face de- tection techniques: a review. Artificial Intelligence Review, 52(2):927–948, 2019

    2019

  20. [28]

    Pruning filters for efficient convnets

    Hao Li, Asim Kadav, Igor Durdanovic, Hanan Samet, and Hans Peter Graf. Pruning filters for efficient convnets. In ICLR, 2017

    2017

  21. [29]

    DSFD: dual shot face detector

    Jian Li, Yabiao Wang, Changan Wang, Ying Tai, Jianjun Qian, Jian Yang, Chengjie Wang, Jilin Li, and Feiyue Huang. DSFD: dual shot face detector. In Proc. CVPR, pages 5060– 5069, 2019

    2019

  22. [30]

    Differentiable transportation pruning

    Yunqiang Li, Jan C van Gemert, Torsten Hoefler, Bert Moons, Evangelos Eleftheriou, and Bram-Ernst Verhoef. Differentiable transportation pruning. In Proc. ICCV, pages 16957–16967, 2023

    2023

  23. [31]

    PruneFaceDet: Pruning lightweight face detection network by sparsity training

    Jingsheng Lin, Xu Zhao, Nanfei Jiang, and Jinqiao Wang. PruneFaceDet: Pruning lightweight face detection network by sparsity training. In Proc. Int. Conf. on Computing and Pattern Recognition, ICCPR ’20, page 181–186, New York, NY , USA, 2021. Association for Computing Machinery

    2021

  24. [32]

    Layer importance estimation with imprinting for neural network quantization

    Hongyang Liu, Sara Elkerdawy, Nilanjan Ray, and Mostafa Elhoushi. Layer importance estimation with imprinting for neural network quantization. In Proc. CVPR, pages 2408– 2417, 2021

    2021

  25. [33]

    SSD: Single shot multibox detector

    Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C Berg. SSD: Single shot multibox detector. In Proc. ECCV, pages 21–37. Springer, 2016

    2016

  26. [34]

    Revisiting token pruning for object detection and instance segmentation

    Yifei Liu, Mathias Gehrig, Nico Messikommer, Marco Can- nici, and Davide Scaramuzza. Revisiting token pruning for object detection and instance segmentation. In Proc. WACV, pages 2658–2668, 2024

    2024

  27. [35]

    ThiNet: A filter level pruning method for deep neural network compression

    Jian-Hao Luo, Jianxin Wu, and Weiyao Lin. ThiNet: A filter level pruning method for deep neural network compression. In Proc. ICCV, Oct 2017

    2017

  28. [36]

    Llm-pruner: On the structural pruning of large language models

    Xinyin Ma, Gongfan Fang, and Xinchao Wang. Llm-pruner: On the structural pruning of large language models. Ad- vances in neural information processing systems, 36:21702– 21720, 2023

    2023

  29. [37]

    Go- ing deeper into face detection: A survey

    Shervin Minaee, Ping Luo, Zhe Lin, and Kevin Bowyer. Go- ing deeper into face detection: A survey. arXiv preprint arXiv:2103.14983, 2021

    2021 arXiv

  30. [38]

    Importance estimation for neural network pruning

    Pavlo Molchanov, Arun Mallya, Stephen Tyree, Iuri Frosio, and Jan Kautz. Importance estimation for neural network pruning. In Proc. CVPR, pages 11264–11272, 2019

    2019

  31. [39]

    Channel-level acceleration of deep face representations

    Adam Polyak and Lior Wolf. Channel-level acceleration of deep face representations. IEEE Access, 3:2163–2175, 2015

    2015

  32. [40]

    YOLO5Face: why reinventing a face detector

    Delong Qi, Weijun Tan, Qi Yao, and Jingfeng Liu. YOLO5Face: why reinventing a face detector. In Proc. ECCV, pages 228–244. Springer, 2022

    2022

  33. [41]

    Automatic enemy detecting defense robot by using face detection technique’

    B Renuka, B Sivaranjani, A Maha Lakshmi, and Dr N Muthukumaran. Automatic enemy detecting defense robot by using face detection technique’. Asian Journal of Applied Science and Technology, 2(2):495–501, 2018

    2018

  34. [42]

    Deep learning-based face detection and recognition on drones

    Mohsen Rostami, Amirhamzeh Farajollahi, and Hashem Parvin. Deep learning-based face detection and recognition on drones. Journal of Ambient Intelligence and Humanized Computing, 15(1):373–387, 2024

    2024

  35. [43]

    Deep feature-based face detection on mobile devices

    Sayantan Sarkar, Vishal M Patel, and Rama Chellappa. Deep feature-based face detection on mobile devices. In Proc. IEEE Int. Conf. on Identity, Security and Behavior Analysis (ISBA), pages 1–8. IEEE, 2016

    2016

  36. [44]

    Taking the human out of the loop: A review of bayesian optimization

    Bobak Shahriari, Kevin Swersky, Ziyu Wang, Ryan P Adams, and Nando De Freitas. Taking the human out of the loop: A review of bayesian optimization. Proceedings of the IEEE, 104(1):148–175, 2015

    2015

  37. [45]

    Prac- tical bayesian optimization of machine learning algorithms

    Jasper Snoek, Hugo Larochelle, and Ryan P Adams. Prac- tical bayesian optimization of machine learning algorithms. Advances in neural information processing systems , 25, 2012

    2012

  38. [46]

    Gaussian process optimization in the ban- dit setting: No regret and experimental design

    Niranjan Srinivas, Andreas Krause, Sham Kakade, and Matthias Seeger. Gaussian process optimization in the ban- dit setting: No regret and experimental design. In In Proc. ICML, pages 1015–1022, 2010

    2010

  39. [47]

    A simple and effective pruning approach for large language models

    Mingjie Sun, Zhuang Liu, Anna Bair, and J Zico Kolter. A simple and effective pruning approach for large language models. arXiv preprint arXiv:2306.11695, 2023

    2023 arXiv

  40. [48]

    Pyra- midbox: A context-assisted single shot face detector

    Xu Tang, Daniel K Du, Zeqiang He, and Jingtuo Liu. Pyra- midbox: A context-assisted single shot face detector. In Proc. ECCV, pages 797–813, 2018

    2018

  41. [49]

    ChipNet: Budget-aware pruning with heaviside con- tinuous approximations

    Rishabh Tiwari, Udbhav Bamba, Arnav Chavan, and Deepak Gupta. ChipNet: Budget-aware pruning with heaviside con- tinuous approximations. In In Proc. ICLR, 2021

    2021

  42. [50]

    Fine-pruning: Joint fine-tuning and compression of a convo- lutional network with bayesian optimization

    Frederick Tung, Srikanth Muralidharan, and Greg Mori. Fine-pruning: Joint fine-tuning and compression of a convo- lutional network with bayesian optimization. arXiv preprint arXiv:1707.09102, 2017

    2017 arXiv

  43. [51]

    Multimodal approach to human-face detection and tracking

    Prahlad Vadakkepat, Peter Lim, Liyanage C De Silva, Liu Jing, and Li Li Ling. Multimodal approach to human-face detection and tracking. IEEE Transactions on Industrial Electronics, 55(3):1385–1393, 2008

    2008

  44. [52]

    Deep rein- forcement learning: A survey

    Xu Wang, Sen Wang, Xingxing Liang, Dawei Zhao, Jincai Huang, Xin Xu, Bin Dai, and Qiguang Miao. Deep rein- forcement learning: A survey. IEEE Transactions on Neural Networks and Learning Systems, 35(4):5064–5078, 2022

    2022

  45. [53]

    Yunet: A tiny millisecond-level face detector

    Wei Wu, Hanyang Peng, and Shiqi Yu. Yunet: A tiny millisecond-level face detector. Machine Intelligence Re- search, 20(5):656–665, 2023

    2023

  46. [54]

    Joint face detection and facial expression recognition with MTCNN

    Jia Xiang and Gengming Zhu. Joint face detection and facial expression recognition with MTCNN. In Proc. Int. Conf. on Information Science and Control Engineering (ICISCE) , pages 424–427, 2017

    2017

  47. [55]

    WIDER FACE: A face detection benchmark

    Shuo Yang, Ping Luo, Chen-Change Loy, and Xiaoou Tang. WIDER FACE: A face detection benchmark. InProc. CVPR, pages 5525–5533, 2016

    2016

  48. [56]

    EXTD: Extremely tiny face detector via iterative filter reuse

    YoungJoon Yoo, Dongyoon Han, and Sangdoo Yun. EXTD: Extremely tiny face detector via iterative filter reuse. arXiv preprint arXiv:1906.06579, 2019

    1906 arXiv

  49. [57]

    Faceboxes: A CPU real-time face de- tector with high accuracy

    Shifeng Zhang, Xiangyu Zhu, Zhen Lei, Hailin Shi, Xiaobo Wang, and Stan Z Li. Faceboxes: A CPU real-time face de- tector with high accuracy. In 2017 IEEE International Joint Conference on Biometrics (IJCB), pages 1–9. IEEE, 2017

    2017

  50. [58]

    S3FD: Single shot scale-invariant face detector

    Shifeng Zhang, Xiangyu Zhu, Zhen Lei, Hailin Shi, Xiaobo Wang, and Stan Z Li. S3FD: Single shot scale-invariant face detector. In Proc. ICCV, pages 192–201, 2017

    2017

  51. [59]

    Real-time multi-scale face detector on em- bedded devices

    Xu Zhao, Xiaoqing Liang, Chaoyang Zhao, Ming Tang, and Jinqiao Wang. Real-time multi-scale face detector on em- bedded devices. Sensors, 19(9):2158, 2019

    2019

  52. [60]

    Real-time multi-scale face detector on em- bedded devices

    Xu Zhao, Xiaoqing Liang, Chaoyang Zhao, Ming Tang, and Jinqiao Wang. Real-time multi-scale face detector on em- bedded devices. Sensors, 2019

    2019

  53. [61]

    Tinaface: Strong but simple baseline for face detection

    Yanjia Zhu, Hongxiang Cai, Shuhan Zhang, Chenhao Wang, and Yichao Xiong. Tinaface: Strong but simple baseline for face detection. arXiv preprint arXiv:2011.13183, 2020

    2011 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.