Pith. sign in

REVIEW 4 major objections 7 minor 47 references

Teacher Supervises Students How to Learn From Partially Labeled Images for Facial Landmark Detection

T0 review · 4 major / 7 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read TS3: a learned teacher filter improves semi-supervised face landmark detection by selecting only reliable pseudo labels, achieving state-of-the-art accuracy with far fewer labels.

desk verdict A solid semi-supervised landmark detection paper with a real teacher-based pseudo-label filtering mechanism, but the evidence for the teacher's transfer is thinner than the claims suggest. read the letter →

arxiv 1908.02116 v3 pith:2XIZM5KS submitted 2019-08-06 cs.CV

Xuanyi Dong , Yi Yang This is my paper · ORCID
classification cs.CV
keywords semi-supervisedlearningfaciallandmarkdetectionpseudolabelsteacher-studentnetworksself-trainingco-trainingheatmapregressionconvolutionalposemachine
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

The paper proposes TS3, an interaction mechanism in which one teacher network scores the quality of pseudo labels produced by two student detectors and keeps only the trustworthy ones for retraining. It claims this learned filter makes semi-supervised facial landmark detection markedly better than prior self-training and co-training approaches, because selection is supervised instead of a heuristic loss. On 300-W, TS3 reports 5.64 NME with 10% labels and 5.03 with 20%, against 6.32 and 5.88 for RCN+. If the claim holds, detectors can approach fully supervised accuracy with a fraction of the landmark annotation cost, and the same teacher-filtering idea transfers to other structured prediction tasks.

What carries the argument

The load-bearing mechanism is the teacher network: a CycleGAN-style discriminator that concatenates the input face image with a student's predicted heatmap and outputs a scalar quality. The teacher is trained on labeled data by L1 regression to match that scalar to the negative squared Frobenius detection loss $\|H^M_i - H^*_i\|_F^2$ (Eq. 4), so a higher quality value is taken to indicate a more reliable pseudo label. This learned filter replaces the heuristic unsupervised loss used by self-paced learning and self-paced co-training baselines. The second mechanism is the dual-student ensemble: two structurally distinct detectors trained on different sample subsets, whose averaged heatmaps define the pseudo label (Eq. 5) and the final prediction.

What would settle it

Train TS3 exactly as described, then on a holdout of unlabeled images with known landmarks compare the teacher's quality scores to the true detection loss of the student pseudo labels; if the correlation is near zero, or if replacing the teacher with random selection at the same ratio keeps the reported NME unchanged, the teacher filter is not doing the work.

Watch

Extended reading notes

Core claim

The central claim is that what limits semi-supervised landmark detection is the reliability of pseudo labels, not their quantity, and that reliability can be learned. TS3 trains a teacher network on the labeled set to regress the negative detection loss of a student's heatmap prediction (Eq. 4), then uses that teacher to score every unlabeled sample and admits only the top $r \times i \times |U|$ samples into the student's retraining set. Two structurally different students, a convolutional pose machine and a stacked hourglass network, alternate so that each generates pseudo labels for the other, and their averaged heatmaps form both the pseudo label (Eq. 5) and the final prediction. The authors report state-of-the-art normalized mean error on 300-W (5.64 and 5.03 with 10% and 20% labels), 1.99 NME on AFLW-Full with 20% labels, and 59.65 AUC on 300-VW category C without using any 300-VW labels.

Load-bearing premise

The whole gain depends on the teacher, trained only on labeled images where student predictions are mostly correct, being able to recognize bad pseudo labels on unlabeled images it has never seen.

Editorial extensions

If this is right

  • At 10% of 300-W labels, TS3 reports 5.64 NME versus 6.32 for RCN+; at 20%, 5.03 versus 5.88, so the learned filter yields lower error with fewer labels than the prior semi-supervised method.
  • With only 20% of AFLW-Full labels, TS3 reaches 1.99 NME, competitive with fully supervised detectors that use all labels.
  • On 300-VW category C, TS3 obtains 59.65 AUC without using any 300-VW labels, above DGCM (59.38) and SBR (59.39) that do use them.
  • In the five-round comparison against self-paced learning baselines, SPL falls into a local trap at round 4 and worsens, while TS3 keeps improving, supporting the claim that teacher-filtered data steers retraining.
  • Because the teacher consumes an image paired with a structured prediction, the same mechanism should apply to segmentation and pose estimation, not just facial landmarks.

Reading between the lines

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

  • A direct test of the weakest point: retrain the teacher on a small validation set drawn from the unlabeled pool with known ground truth, as the authors suggest, and measure whether teacher quality scores correlate with true detection loss; if not, the filter is no better than random.
  • The reported 64x64 resolution limit suggests the teacher's quality function is local; a patch-based or multi-scale teacher might extend the mechanism to high-resolution segmentation while keeping the same training signal.
  • The dual-student ensemble and the teacher filter are confounded in the headline numbers; ablating each separately (e.g., teacher with a single student, ensemble without teacher) would tell which component carries the gain.
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

4 major / 7 minor

Summary. The paper studies semi-supervised facial landmark detection and proposes TS3, an interaction mechanism between one teacher network and two student detectors (CPM and stacked hourglass). The students generate pseudo-label heatmaps for unlabeled images; the teacher, trained on labeled images to regress the negative detection loss of student predictions, scores each pseudo-labeled sample, and the top-ranked samples are progressively added to the training set. The final prediction is an ensemble of the two students. Experiments on 300-W, AFLW, and 300-VW report improved NME/AUC over the semi-supervised baseline RCN+ and over loss-based selection strategies SPL and SPaCo, and the authors claim state-of-the-art semi-supervised performance.

Significance. If the central mechanism works as claimed, the paper makes a useful contribution: a learned, data-driven quality filter for pseudo-labels in a structured-prediction task, with public code and results on three standard benchmarks. The ablations partially isolate the two claimed components (teacher-based selection and dual-student interaction), and the comparison against external baselines means the empirical claims are not circular. However, the core assumption that a teacher trained on nearly correct labeled-data predictions can rank the much noisier pseudo-labels on unlabeled data is not directly verified, and the reported results lack variance information. The current evidence supports competitiveness but falls short of fully establishing the stated mechanism and the abstract's 30%-label claim.

major comments (4)
  1. [§3.2, Eq. (4), Algorithm 1 steps 8 and 13] The teacher is trained only on the labeled set L, where student predictions are close to the ground truth, so the regression target q = -||H^M_i - H*_i||^2_F is concentrated near zero and the teacher rarely sees bad pseudo-labels. Section 3.4 explicitly concedes this over-fitting risk: the teacher 'might only learn what a good pseudo labeled sample is, but overlook what a bad one is.' The central claim that teacher filtering is more reliable than loss-based selection (SPL/SPaCo) therefore rests on an unverified transfer assumption. Please provide direct evidence that teacher scores on U correlate with actual detection error, for example a rank-correlation analysis on a held-out subset or a comparison of the error of selected versus rejected pseudo-labels, and supplement Figure 4 with variance across the three repeats. Without this, the advantage over SPL/SPaCo could arise from near-random selection or from input-difficulty bias rather than from the teacher's ability to identify incorrect pseudo-labels.
  2. [Abstract and Section 1] The abstract and introduction state that using only 30% labels the method achieves competitive results with full supervision on 300-W and AFLW, but no 30%-label experiment appears in Tables 1 or 2, which report only 100%, 20%, 10%, and 5% ratios. This is a missing support for a specific claimed result. Either add the 30% experiments or remove/qualify the claim.
  3. [§4.2 and Tables 1, 2, 4; Figure 4] The paper states that each experiment is repeated three times and the mean is reported, yet no standard deviations, confidence intervals, or per-repeat values are given. Tables 1, 2, and 4 show single numbers, and Figure 4 shows single trajectories without error bars. Because L and U are randomly sampled and the teacher and students are randomly initialized, the differences used to support the state-of-the-art claim (for example, 5.64 vs. 6.32 NME at 10% labels in Table 1) could be within run-to-run noise. Please report variance or a significance measure for the main comparisons and for the Figure 4 curves.
  4. [§4.4, Table 4] The ablation on the 'effect of the interaction between students' compares single-student CPM and HG rows with the TS3 row, but the TS3 row uses the Eq. (5) ensemble of two detectors at evaluation time, whereas the single-student rows use only one detector. This conflates the mutual-training interaction with test-time ensembling, so the reported 30% and 9% improvements could be due to ensemble averaging alone. Please add a control in which two detectors are trained independently without interaction and then ensembled by Eq. (5), or report the individual student errors inside TS3, to isolate the contribution of the student-student interaction.
minor comments (7)
  1. [§4.2] The selection ratio r and maximum step S are said to be set 'based on cross-validation,' but no details are given about the validation split or the validation criterion. Please specify what was validated and on which data.
  2. [Algorithm 1, lines 7 and 12] The symbol '⊿' and the wording 'update the first student' are unclear; also, because Eq. (5) uses both f1 and f2, the temporal ordering of the two students' predictions when updating a single student should be stated explicitly.
  3. [Eq. (4)] The notation is inconsistent: Eq. (3) defines q_i while Eq. (4) writes q without an index, and it is not stated whether the quality scalar is computed from the last stage only or from all M stages. Please clarify.
  4. [Table 1] RCN+ is the key semi-supervised baseline but its citation appears only in the surrounding text; adding [16] in the table caption or row would improve readability.
  5. [Table 3] Please clarify whether DGCM and SBR use the same 300-W training labels in addition to the 300-VW labels, and how the 300-VW unlabeled training frames are selected and processed for TS3, since the comparison depends on the exact data split.
  6. [Figure 5 caption] The caption's references to 'the third line' and 'the fifth line' do not match the row descriptions in the text (second row ground truth heatmaps, third row ground truth landmarks, fourth row predicted heatmaps, fifth row predicted coordinates). Please make the caption consistent.
  7. [Section 3.4] The statement that the teacher 'will fail' at 256x256 input is presented without experiments; please mark it as a conjecture or provide supporting evidence.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: TS3's core claim is evaluated on external benchmarks, and the teacher's quality regression is a learned surrogate for detection loss rather than a re-use of the target result.

full rationale

The paper's central claim is that a teacher network can rank student-generated pseudo labels on unlabeled images and thereby improve semi-supervised facial landmark detection, as measured on the external 300-W, AFLW, and 300-VW benchmarks. This claim is not circular: the teacher is trained by Eq. (4) to regress a scalar quality to the negative detection loss on labeled data, and its ranking on unlabeled data is an empirical transfer that can fail. The paper itself acknowledges the key risk in Section 3.4: 'the teacher's judgment could suffer from the over-fitting problem... it might only learn what a good pseudo labeled sample is, but overlook what a bad one is.' This admission is a limitation and a correctness risk, not evidence of circularity, because the final reported numbers come from held-out test sets rather than from the teacher's own training objective. No load-bearing step reduces to a fitted parameter or to a self-citation by construction. The self-citations in the paper ([9], [11], [12], [28]) appear as baselines, related work, or supervised counterparts; they do not supply an unverified premise on which the main result depends. The teacher's quality target being defined as negative detection loss is a design choice, but using a learned predictor of that target to select pseudo-labeled samples is a meaningful algorithmic proposal with independent benchmark evaluation. Therefore the circularity score is 0.

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

The core mechanism rests on the teacher's ability to generalize from labeled to unlabeled data, which the authors themselves flag as an overfitting risk (Sec 3.4). The remaining axioms are standard domain assumptions for semi-supervised landmark detection: detection loss is a meaningful quality signal, the two chosen architectures complement each other, and the linear selection schedule is a pragmatic choice. No new physical entities are introduced.

free parameters (3)
  • selection ratio r = 0.1
    Chosen by cross-validation (Sec 4.2). Controls how many pseudo-labeled samples are added per round in Algorithm 1; the central results depend on this value.
  • maximum step S = 6
    Chosen by cross-validation (Sec 4.2). Number of teacher-student interaction rounds; stopping earlier or later changes final accuracy.
  • input resolution = 64x64
    Paper states 64x64 is chosen to ease teacher training (footnote 2 and Sec 3.4); higher resolutions make the teacher fail, limiting the method.
assumptions (4)
  • domain assumption The quality function learned on labeled data transfers to unlabeled data.
    Teacher g is trained on L via Eq. (4) and then applied to U in Algorithm 1; Section 3.4 admits overfitting risk because L contains mostly well-aligned pseudo-labels.
  • domain assumption Low detection loss of a pseudo-label correlates with landmark accuracy and with usefulness for retraining.
    The teacher's target is the negative detection loss (Eq. 4); the whole filtering mechanism assumes that samples with low predicted loss are the best training data, which is plausible but not proven.
  • domain assumption The two students are complementary, so averaging their heatmaps (Eq. 5) yields better pseudo-labels and final predictions than either alone.
    Supported by Table 4 ablation, but the choice of CPM+HG is one instantiation; the gain could depend on this specific pair.
  • ad hoc to paper The number of selected pseudo-labeled samples should grow linearly with round index (r x i x |U|).
    Algorithm 1 increases the selected set each round; this schedule is a design choice without theoretical justification and affects the learning dynamics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Teacher Supervises Students How to Learn From Partially Labeled Images for Facial Landmark Detection." pith.science (2026). https://pith.science/paper/2XIZM5KS

@misc{pith2026190802116,
  author       = {Pith},
  title        = {Pith review of: Teacher Supervises Students How to Learn From Partially Labeled Images for Facial Landmark Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2XIZM5KS}},
  note         = {Machine review of arXiv:1908.02116}
}
read the original abstract

Facial landmark detection aims to localize the anatomically defined points of human faces. In this paper, we study facial landmark detection from partially labeled facial images. A typical approach is to (1) train a detector on the labeled images; (2) generate new training samples using this detector's prediction as pseudo labels of unlabeled images; (3) retrain the detector on the labeled samples and partial pseudo labeled samples. In this way, the detector can learn from both labeled and unlabeled data to become robust. In this paper, we propose an interaction mechanism between a teacher and two students to generate more reliable pseudo labels for unlabeled data, which are beneficial to semi-supervised facial landmark detection. Specifically, the two students are instantiated as dual detectors. The teacher learns to judge the quality of the pseudo labels generated by the students and filter out unqualified samples before the retraining stage. In this way, the student detectors get feedback from their teacher and are retrained by premium data generated by itself. Since the two students are trained by different samples, a combination of their predictions will be more robust as the final prediction compared to either prediction. Extensive experiments on 300-W and AFLW benchmarks show that the interactions between teacher and students contribute to better utilization of the unlabeled data and achieves state-of-the-art performance.

Figures

Figures reproduced from arXiv: 1908.02116 by the authors.

Figure 1
Figure 1. The interaction mechanism between teacher and stu [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. A brief overview of the structure between the two stu [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The illustration of our teacher network. The input of the teacher is the concatenation of the original RGB face image and the heatmap (pseudo label) predicted by the student detector. The output of teacher is a scalar, representing the quality of the input pseudo labeled face image. During training, we can calculate a detection loss using the ideal heatmap and the predicted heatmap. The teacher aims to fit the negat… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: We compare three different algorithms, which can train [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Qualitative results on images in the 300-W test set. We train our TS3 with 314 labeled facial images and 2834 unla￾beled facial images in the 300-W training set. can be readily applied to multiple students without introduc￾ing additional hyper-parameters. In contrast, …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 42 canonical work pages

  1. [1]

    Learning with pseudo-ensembles

    Phil Bachman, Ouais Alsharif, and Doina Precup. Learning with pseudo-ensembles. In NeurIPS, 2014

    work page 2014

  2. [2]

    Curriculum learning

    Yoshua Bengio, J ´erˆome Louradour, Ronan Collobert, and Ja- son Weston. Curriculum learning. In ICML, 2009

    work page 2009

  3. [3]

    Face recognition based on fitting a 3d morphable model

    V olker Blanz and Thomas Vetter. Face recognition based on fitting a 3d morphable model. IEEE TPAMI, 2003

    work page 2003

  4. [4]

    Combining labeled and un- labeled data with co-training

    Avrim Blum and Tom Mitchell. Combining labeled and un- labeled data with co-training. In CLT, 1998

    work page 1998

  5. [5]

    Convolutional aggregation of local evidence for large pose face alignment

    Adrian Bulat and Georgios Tzimiropoulos. Convolutional aggregation of local evidence for large pose face alignment. In BMVC, 2016

    work page 2016

  6. [6]

    How far are we from solving the 2D & 3D face alignment problem? (and a dataset of 230,000 3D facial landmarks)

    Adrian Bulat and Georgios Tzimiropoulos. How far are we from solving the 2D & 3D face alignment problem? (and a dataset of 230,000 3D facial landmarks). In ICCV, 2017

    work page 2017

  7. [7]

    Face alignment by explicit shape regression

    Xudong Cao, Yichen Wei, Fang Wen, and Jian Sun. Face alignment by explicit shape regression. IJCV, 2014

    work page 2014

  8. [8]

    Semi-supervised learning

    Olivier Chapelle, Bernhard Sch ¨olkopf, and Alexander Zien. Semi-supervised learning. MIT press Cambridge, 2006

    work page 2006

Show all 47 references
  1. [9]

    Style aggregated network for facial landmark detection

    Xuanyi Dong, Yan Yan, Wanli Ouyang, and Yi Yang. Style aggregated network for facial landmark detection. In CVPR, 2018

    2018

  2. [10]

    Network pruning via transformable architecture search

    Xuanyi Dong and Yi Yang. Network pruning via transformable architecture search. arXiv preprint arXiv:1905.09717, 2019

    1905 arXiv

  3. [11]

    Supervision-by-Registration: An unsupervised approach to improve the precision of facial landmark detectors

    Xuanyi Dong, Shoou-I Yu, Xinshuo Weng, Shih-En Wei, Yi Yang, and Yaser Sheikh. Supervision-by-Registration: An unsupervised approach to improve the precision of facial landmark detectors. In CVPR, 2018

    2018

  4. [12]

    Few-example object detection with model communi- cation

    Xuanyi Dong, Liang Zheng, Fan Ma, Yi Yang, and Deyu Meng. Few-example object detection with model communi- cation. IEEE TPAMI, 2018

    2018

  5. [13]

    Learning to teach

    Yang Fan, Fei Tian, Tao Qin, Xiang-Yang Li, and Tie-Yan Liu. Learning to teach. In ICLR, 2018

    2018

  6. [14]

    Generative adversarial nets

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In NeurIPS, 2014

    2014

  7. [15]

    Distilling the knowledge in a neural network

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the knowledge in a neural network. InNeurIPS Workshop, 2014

    2014

  8. [16]

    Improving landmark localization with semi-supervised learning

    Sina Honari, Pavlo Molchanov, Stephen Tyree, Pascal Vin- cent, Christopher Pal, and Jan Kautz. Improving landmark localization with semi-supervised learning. In CVPR, 2018

    2018

  9. [17]

    Self-paced curriculum learning

    Lu Jiang, Deyu Meng, Qian Zhao, Shiguang Shan, and Alexander G Hauptmann. Self-paced curriculum learning. In AAAI, 2015

    2015

  10. [18]

    MentorNet: Learning data-driven curriculum for very deep neural networks on corrupted labels

    Lu Jiang, Zhengyuan Zhou, Thomas Leung, Li-Jia Li, and Li Fei-Fei. MentorNet: Learning data-driven curriculum for very deep neural networks on corrupted labels. In ICML, 2018

    2018

  11. [19]

    Pose- invariant face alignment with a single cnn

    Amin Jourabloo, Xiaoming Liu, Mao Ye, and Liu Ren. Pose- invariant face alignment with a single cnn. In ICCV, 2017

    2017

  12. [20]

    Synergy between face alignment and tracking via discriminative global consensus optimization

    Muhammad Haris Khan, John McDonagh, and Georgios Tz- imiropoulos. Synergy between face alignment and tracking via discriminative global consensus optimization. In ICCV, 2017

    2017

  13. [21]

    Annotated facial landmarks in the wild: A large- scale, real-world database for facial landmark localization

    Martin Koestinger, Paul Wohlhart, Peter M Roth, and Horst Bischof. Annotated facial landmarks in the wild: A large- scale, real-world database for facial landmark localization. In ICCV Workshop, 2011

    2011

  14. [22]

    Disentangling 3D pose in a dendritic CNN for unconstrained 2D face alignment

    Amit Kumar and Rama Chellappa. Disentangling 3D pose in a dendritic CNN for unconstrained 2D face alignment. In CVPR, 2018

    2018

  15. [23]

    Self- paced learning for latent variable models

    M Pawan Kumar, Benjamin Packer, and Daphne Koller. Self- paced learning for latent variable models. In NeurIPS, 2010

    2010

  16. [24]

    Teacher and student joint learning for compact facial landmark detection network

    Hong Joo Lee, Wissam J Baddar, Hak Gu Kim, Seong Tae Kim, and Yong Man Ro. Teacher and student joint learning for compact facial landmark detection network. In ICMM, 2018

    2018

  17. [25]

    Prototype propagation networks (PPN) for weakly-supervised few-shot learning on category graph

    Lu Liu, Tianyi Zhou, Guodong Long, Jing Jiang, Lina Yao, and Chengqi Zhang. Prototype propagation networks (PPN) for weakly-supervised few-shot learning on category graph. In IJCAI, 2019

    2019

  18. [26]

    Exploring disentangled feature representa- tion beyond face identification

    Yu Liu, Fangyin Wei, Jing Shao, Lu Sheng, Junjie Yan, and Xiaogang Wang. Exploring disentangled feature representa- tion beyond face identification. In CVPR, 2018

    2018

  19. [27]

    A deep regression architecture with two-stage reinitialization for high performance facial landmark detec- tion

    Jiangjing Lv, Xiaohu Shao, Junliang Xing, Cheng Cheng, and Xi Zhou. A deep regression architecture with two-stage reinitialization for high performance facial landmark detec- tion. In CVPR, 2017

    2017

  20. [28]

    Self-paced co-training

    Fan Ma, Deyu Meng, Qi Xie, Zina Li, and Xuanyi Dong. Self-paced co-training. In ICML, 2017

    2017

  21. [29]

    Direct shape regression net- works for end-to-end face alignment

    Xin Miao, Xiantong Zhen, Xianglong Liu, Cheng Deng, Vas- silis Athitsos, and Heng Huang. Direct shape regression net- works for end-to-end face alignment. In CVPR, 2018

    2018

  22. [30]

    Stacked hour- glass networks for human pose estimation

    Alejandro Newell, Kaiyu Yang, and Jia Deng. Stacked hour- glass networks for human pose estimation. In ECCV, 2016

    2016

  23. [31]

    Data distillation: Towards omni- supervised learning

    Ilija Radosavovic, Piotr Doll ´ar, Ross Girshick, Georgia Gkioxari, and Kaiming He. Data distillation: Towards omni- supervised learning. In CVPR, 2018

    2018

  24. [32]

    Hy- perface: A deep multi-task learning framework for face de- tection, landmark localization, pose estimation, and gender recognition

    Rajeev Ranjan, Vishal M Patel, and Rama Chellappa. Hy- perface: A deep multi-task learning framework for face de- tection, landmark localization, pose estimation, and gender recognition. IEEE TPAMI, 2019

    2019

  25. [33]

    Learning to reweight examples for robust deep learning

    Mengye Ren, Wenyuan Zeng, Bin Yang, and Raquel Urta- sun. Learning to reweight examples for robust deep learning. In ICML, 2018

    2018

  26. [34]

    Face alignment via regressing local binary features

    Shaoqing Ren, Xudong Cao, Yichen Wei, and Jian Sun. Face alignment via regressing local binary features. IEEE TIP, 2016

    2016

  27. [35]

    300 faces in-the-wild challenge: The first facial landmark localization challenge

    Christos Sagonas, Georgios Tzimiropoulos, Stefanos Zafeiriou, and Maja Pantic. 300 faces in-the-wild challenge: The first facial landmark localization challenge. In ICCV Workshop, 2013

    2013

  28. [36]

    The first facial landmark tracking in-the-wild challenge: Benchmark and results

    Jie Shen, Stefanos Zafeiriou, Grigoris G Chrysos, Jean Kos- saifi, Georgios Tzimiropoulos, and Maja Pantic. The first facial landmark tracking in-the-wild challenge: Benchmark and results. In ICCV Workshop, 2015

    2015

  29. [37]

    Quantized densely connected u-nets for efficient landmark localization

    Zhiqiang Tang, Xi Peng, Shijie Geng, Lingfei Wu, Shaoting Zhang, and Dimitris Metaxas. Quantized densely connected u-nets for efficient landmark localization. In ECCV, 2018

    2018

  30. [38]

    Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results

    Antti Tarvainen and Harri Valpola. Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results. In NeurIPS, 2017. 9

    2017

  31. [39]

    Face2face: Real-time face capture and reenactment of rgb videos

    Justus Thies, Michael Zollhofer, Marc Stamminger, Chris- tian Theobalt, and Matthias Nießner. Face2face: Real-time face capture and reenactment of rgb videos. In CVPR, 2016

    2016

  32. [40]

    Mnemonic descent method: A recurrent process applied for end-to-end face alignment

    George Trigeorgis, Patrick Snape, Mihalis A Nico- laou, Epameinondas Antonakos, and Stefanos Zafeiriou. Mnemonic descent method: A recurrent process applied for end-to-end face alignment. In CVPR, 2016

    2016

  33. [41]

    Convolutional pose machines

    Shih-En Wei, Varun Ramakrishna, Takeo Kanade, and Yaser Sheikh. Convolutional pose machines. In CVPR, 2016

    2016

  34. [42]

    Facial landmark detection with tweaked convolutional neural networks

    Yue Wu, Tal Hassner, KangGeon Kim, Gerard Medioni, and Prem Natarajan. Facial landmark detection with tweaked convolutional neural networks. IEEE TPAMI, 2017

    2017

  35. [43]

    Recurrent 3d- 2d dual learning for large-pose facial landmark detection

    Shengtao Xiao, Jiashi Feng, Luoqi Liu, Xuecheng Nie, Wei Wang, Shuicheng Yan, and Ashraf Kassim. Recurrent 3d- 2d dual learning for large-pose facial landmark detection. In CVPR, 2017

    2017

  36. [44]

    Supervised descent method and its applications to face alignment

    Xuehan Xiong and Fernando De la Torre. Supervised descent method and its applications to face alignment. In CVPR, 2013

    2013

  37. [45]

    AutoLoss: Learning discrete schedules for alternate optimization

    Haowen Xu, Hao Zhang, Zhiting Hu, Xiaodan Liang, Ruslan Salakhutdinov, and Eric Xing. AutoLoss: Learning discrete schedules for alternate optimization. In ICLR, 2019

    2019

  38. [46]

    Unpaired image-to-image translation using cycle- consistent adversarial networks

    Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A Efros. Unpaired image-to-image translation using cycle- consistent adversarial networks. In ICCV, 2017

    2017

  39. [47]

    Unconstrained face alignment via cascaded compo- sitional learning

    Shizhan Zhu, Cheng Li, Chen-Change Loy, and Xiaoou Tang. Unconstrained face alignment via cascaded compo- sitional learning. In CVPR, 2016. 10

    2016

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.