arxiv: 2605.14689 · v1 · submitted 2026-05-14 · 💻 cs.CV

Recognition: 2 theorem links

· Lean Theorem

Are Candidate Models Really Needed for Active Learning?

Harshini Mridula Mohan , Maanya Manjunath , Vipul Arya , S.H. Shabbeer Basha , Nitin Cheekatla

Authors on Pith no claims yet

Pith reviewed 2026-05-15 05:15 UTC · model grok-4.3

classification 💻 cs.CV

keywords active learningrandom initializationCNNtransformersconfidence samplinglow confidencesample selectioncomputer vision

0 comments

The pith

Randomly initialized CNNs and transformers can guide active learning sample selection without any initial candidate models.

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

The paper tests whether active learning needs pre-trained candidate models by running CNNs and transformers from random weights instead. It evaluates three confidence-based selection rules—high confidence, low confidence, and a switch from high to low—and finds that low-confidence sampling works as well as standard active-learning pipelines on multiple vision datasets. Removing the candidate-model step cuts the usual two-stage training loop down to a single process that still reduces labeling effort. The results hold across architectures and datasets, showing that useful uncertainty signals appear even before any supervised training begins. This removes a major setup cost while preserving the accuracy gains of active learning.

Core claim

CNNs and transformers with randomly initialized weights produce confidence scores that are informative enough to select which samples to label next; low-confidence sampling among these scores yields final model performance comparable to conventional active-learning methods that first train a candidate model on an initial labeled pool.

What carries the argument

Confidence-based sampling (high-confidence, low-confidence, and hybrid HCLC) applied directly to the softmax outputs of randomly initialized CNNs and transformers.

If this is right

Active learning pipelines can begin without any preliminary training round.
Low-confidence sampling alone suffices as the selection rule for the entire labeling process.
Both convolutional and transformer architectures support the random-initialization approach.
Total wall-clock time for active learning drops because the candidate-model training step disappears.
The method applies across standard computer-vision benchmarks without architecture-specific tuning.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same random-init confidence signal might work for non-vision tasks where pre-training is expensive.
If low-confidence selection succeeds early, it could reduce the total number of labeling rounds needed.
The approach invites direct comparison with uncertainty methods that do require pre-training to see where the random-init shortcut breaks.
Resource-limited settings could adopt active learning immediately without waiting for a strong base model.

Load-bearing premise

The assumption that the raw confidence numbers coming from a network with random weights actually point to the most useful samples to label.

What would settle it

Run the same labeling budget with uniform random selection instead of low-confidence selection from random-init models; if the final test accuracy is statistically identical, the claim is falsified.

Figures

Figures reproduced from arXiv: 2605.14689 by Harshini Mridula Mohan, Maanya Manjunath, Nitin Cheekatla, S.H. Shabbeer Basha, Vipul Arya.

**Figure 2.** Figure 2: Comparison of model performances at different proportions of training data. CIFAR [PITH_FULL_IMAGE:figures/full_fig_p020_2.png] view at source ↗

**Figure 3.** Figure 3: Comparison of active learning methods on the SVHN dataset with ResNet-18. [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗

**Figure 4.** Figure 4: Comparison of various developed sampling methods results on CIFAR-10 using [PITH_FULL_IMAGE:figures/full_fig_p024_4.png] view at source ↗

**Figure 5.** Figure 5: Comparison of active learning results for DinoV2 using pre-trained vs. DinoV2 with [PITH_FULL_IMAGE:figures/full_fig_p027_5.png] view at source ↗

read the original abstract

Deep learning has profoundly impacted domains such as computer vision and natural language processing by uncovering complex patterns in vast datasets. However, the reliance on extensive labeled data poses significant challenges, including resource constraints and annotation errors, particularly in training Convolutional Neural Networks (CNNs) and transformers due to a larger number of parameters. Active learning offers a promising solution to reduce labeling burdens by strategically selecting the most informative samples for annotation. However, the current active learning frameworks are time-intensive which select the samples iteratively with the help of initial candidate models. This study investigates the feasibility of using CNNs and transformers with randomly initialized weights, eliminating the need for initial candidate models while achieving results comparable to active learning frameworks that depend on such candidate models. We evaluate three confidence-based sampling strategies: high confidence (HC), low confidence (LC), and a combination of high confidence in the early stages of training and low confidence at later stages of training (HCLC). Among these, mostly LC demonstrated the best performance in our experiments, showcasing its effectiveness as an active learning strategy without the need for candidate models. Further, extensive experiments verify the robustness of the proposed active learning methods. By challenging traditional frameworks, the proposed work introduces a streamlined approach to active learning, advancing efficiency and flexibility across diverse datasets and domains.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Random initialization can stand in for trained candidate models in confidence-based active learning, but only if the experiments show it beats plain random selection.

read the letter

The main point is that you can skip training any initial candidate models and still run low-confidence sampling on randomly initialized CNNs and transformers. The authors test high-confidence, low-confidence, and a hybrid HCLC strategy, report that low confidence performs best, and claim the results stay comparable to the usual active learning setups that start with trained models. This removes the upfront training loop and cuts compute at the start of the process.

Referee Report

2 major / 1 minor

Summary. The paper claims that active learning for CNNs and transformers can be performed effectively using randomly initialized weights, without any initial candidate models, by applying confidence-based sampling strategies (high confidence HC, low confidence LC, and hybrid HCLC). It asserts that LC yields the best performance among these and achieves results comparable to traditional AL frameworks that rely on pre-trained or candidate models, with extensive experiments verifying robustness across datasets.

Significance. If the central empirical claim holds with proper controls, the work would offer a meaningful simplification of active learning pipelines by removing the need for candidate model training, potentially improving efficiency in resource-constrained settings. However, the significance is limited by the absence of clear quantitative evidence in the provided abstract and the unresolved question of whether random-initialization confidence scores carry genuine informativeness beyond random selection.

major comments (2)

[Experiments] Experiments section: the central claim that LC selection with random-init models outperforms or matches traditional AL requires a random-sampling baseline; without it, any reported gains could be consistent with the null that the ordering induced by untrained softmax outputs is uncorrelated with true informativeness, reducing the method to non-adaptive ranking.
[Results] Results and evaluation: the abstract states 'comparable results' and 'best performance' for LC but provides no numerical metrics, dataset sizes, number of runs, or statistical significance tests; these details are load-bearing for the claim that random-init AL eliminates the need for candidate models.

minor comments (1)

[Method] Clarify the precise definition of 'randomly initialized weights' (e.g., whether any pre-training on unrelated data is used) and how the final model is trained after selection.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the major comments point by point below, agreeing where revisions are needed to strengthen the empirical claims.

read point-by-point responses

Referee: Experiments section: the central claim that LC selection with random-init models outperforms or matches traditional AL requires a random-sampling baseline; without it, any reported gains could be consistent with the null that the ordering induced by untrained softmax outputs is uncorrelated with true informativeness, reducing the method to non-adaptive ranking.

Authors: We agree that a random-sampling baseline is required to establish that the confidence-based ordering is informative rather than equivalent to random selection. In the revised manuscript we will add random sampling as an explicit baseline in all experiments, reporting accuracy curves and final performance deltas versus LC (and HC/HCLC) on every dataset. This will allow direct verification that the untrained softmax scores induce a non-random ranking correlated with true informativeness. revision: yes
Referee: Results and evaluation: the abstract states 'comparable results' and 'best performance' for LC but provides no numerical metrics, dataset sizes, number of runs, or statistical significance tests; these details are load-bearing for the claim that random-init AL eliminates the need for candidate models.

Authors: We will revise the abstract to include concrete numerical results (e.g., final accuracies or AUCs for LC versus traditional candidate-model AL and versus random sampling), dataset sizes, number of independent runs, and a statement that statistical significance was assessed via paired t-tests or Wilcoxon tests. The full tables and significance values already appear in the experiments section; the abstract update will make these load-bearing details immediately visible while preserving brevity. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical comparison with no derivation chain

full rationale

The paper conducts an experimental study evaluating confidence-based active learning strategies (HC, LC, HCLC) on randomly initialized CNNs and transformers across datasets. No mathematical derivations, uniqueness theorems, fitted parameters renamed as predictions, or self-citation load-bearing steps appear in the provided text. All claims rest on direct empirical comparisons to baselines, with no reduction of results to inputs by construction. The approach is self-contained as a standard empirical ablation.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review; no explicit free parameters or invented entities are mentioned. The central claim rests on the domain assumption that random-initialization confidence scores remain useful for selection.

axioms (1)

domain assumption Confidence estimates from randomly initialized neural networks are informative enough for active learning sample selection.
This premise is required for the claim that candidate models can be eliminated.

pith-pipeline@v0.9.0 · 5541 in / 1204 out tokens · 75728 ms · 2026-05-15T05:15:53.247742+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

This study investigates the feasibility of using CNNs and transformers with randomly initialized weights, eliminating the need for initial candidate models while achieving results comparable to active learning frameworks that depend on such candidate models. We evaluate three confidence-based sampling strategies: high confidence (HC), low confidence (LC), and a combination... mostly LC demonstrated the best performance
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

The Lottery Ticket Hypothesis... randomly initialized neural networks there exist sub-networks... capable of achieving comparable performance

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

180 extracted references · 180 canonical work pages · 16 internal anchors

[1]

Journal of Visual Communication and Image Representation , pages=

Gradient-based conditional generative adversarial network for non-uniform blind deblurring via DenseResNet , author=. Journal of Visual Communication and Image Representation , pages=. 2020 , publisher=

work page 2020
[2]

NeurIPS , year=

Dual Swap Disentangling , author=. NeurIPS , year=

work page
[3]

International Conference on Learning Representations (ICLR) , year=

Very Deep Convolutional Networks for Large-Scale Image Recognition , author=. International Conference on Learning Representations (ICLR) , year=

work page
[4]

European Conference on Computer Vision (ECCV) , year=

SSD: Single Shot MultiBox Detector , author=. European Conference on Computer Vision (ECCV) , year=

work page
[5]

arXiv preprint arXiv:1712.00960 , year=

FSSD: Feature Fusion Single Shot Multibox Detector , author=. arXiv preprint arXiv:1712.00960 , year=

work page arXiv
[6]

Light-Head R-CNN: In Defense of Two-Stage Object Detector

Feature-Fused SSD: Fast Detection for Small Objects , author=. arXiv preprint arXiv:1711.07264 , year=

work page internal anchor Pith review Pith/arXiv arXiv
[7]

Competing phases and topological excitations of spin-one pyrochlore antiferromagnets

Concatenated Feature Maps for SSD , author=. arXiv preprint arXiv:1712.00740 , year=

work page internal anchor Pith review Pith/arXiv arXiv
[8]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

Plug-and-Play Active Learning for Object Detection , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

work page
[9]

Journal of Visual Communication and Image Representation , volume=

Improved-StoryGAN for sequential images visualization , author=. Journal of Visual Communication and Image Representation , volume=. 2020 , publisher=

work page 2020
[10]

Journal of Visual Communication and Image Representation , volume=

AP-GAN: Predicting skeletal activity to improve early activity recognition , author=. Journal of Visual Communication and Image Representation , volume=. 2020 , publisher=

work page 2020
[11]

Journal of Visual Communication and Image Representation , pages=

Conditional Generative Adversarial Network for EEG-based Emotion Fine-grained Estimation and Visualization , author=. Journal of Visual Communication and Image Representation , pages=. 2020 , publisher=

work page 2020
[12]

Journal of Visual Communication and Image Representation , volume=

Semi-supervised cross-modal representation learning with GAN-based Asymmetric Transfer Network , author=. Journal of Visual Communication and Image Representation , volume=. 2020 , publisher=

work page 2020
[13]

Advances in neural information processing systems , pages=

Generative adversarial nets , author=. Advances in neural information processing systems , pages=

work page
[14]

IEEE Conference on Computer Vision and Pattern Recognition , pages=

Image-to-Image Translation with Conditional Adversarial Networks , author=. IEEE Conference on Computer Vision and Pattern Recognition , pages=

work page
[15]

IEEE International Conference on Computer Vision , pages=

Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks , author=. IEEE International Conference on Computer Vision , pages=

work page
[16]

IEEE International Conference on Computer Vision , pages=

DualGAN: Unsupervised Dual Learning for Image-to-Image Translation , author=. IEEE International Conference on Computer Vision , pages=

work page
[17]

IEEE International Conference on Automatic Face & Gesture Recognition , pages=

High-quality facial photo-sketch synthesis using multi-adversarial networks , author=. IEEE International Conference on Automatic Face & Gesture Recognition , pages=

work page
[18]

ACM International Conference on Multimedia Retrieval , pages=

End-to-end photo-sketch generation via fully convolutional representation learning , author=. ACM International Conference on Multimedia Retrieval , pages=

work page
[19]

Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

Fully convolutional networks for semantic segmentation , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

work page
[20]

Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

Image style transfer using convolutional neural networks , author=. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

work page
[21]

IEEE Transactions on Pattern Analysis & Machine Intelligence , number=

Face photo-sketch synthesis and recognition , author=. IEEE Transactions on Pattern Analysis & Machine Intelligence , number=

work page
[22]

European Conference on Computer Vision , year=

Colorful Image Colorization , author=. European Conference on Computer Vision , year=

work page
[23]

IEEE International Conference on Computer Vision , year =

Cheng, Zezhou and Yang, Qingxiong and Sheng, Bin , title =. IEEE International Conference on Computer Vision , year =

work page
[24]

How to measure uncertainty in uncertainty sampling for active learning , journal =

Vu-Linh Nguyen and Mohammad Hossein Shaker and Eyke H. How to measure uncertainty in uncertainty sampling for active learning , journal =. 2022 , doi =

work page 2022
[25]

Pattern Recognition , volume =

Edwin Lughofer , title =. Pattern Recognition , volume =. 2012 , url =

work page 2012
[26]

Gupta and Xiaojiang Chen and Xin Wang , title =

Pengzhen Ren and Yun Xiao and Xiaojun Chang and Po-Yao Huang and Zhihui Li and Brij B. Gupta and Xiaojiang Chen and Xin Wang , title =. arXiv preprint , volume =. 2021 , month =

work page 2021
[27]

European Conference on Computer Vision , pages=

Learning representations for automatic colorization , author=. European Conference on Computer Vision , pages=. 2016 , organization=

work page 2016
[28]

Spatial Pattern Templates for Recognition of Objects with Regular Structure , booktitle =

Radim Tyle. Spatial Pattern Templates for Recognition of Objects with Regular Structure , booktitle =

work page
[29]

IEEE Global Conference on Signal and Information Processing , pages=

Deep learning based image super-resolution with coupled backpropagation , author=. IEEE Global Conference on Signal and Information Processing , pages=

work page
[30]

IEEE International Conference on Signal Processing , pages=

Single image super-resolution based on deep learning and gradient transformation , author=. IEEE International Conference on Signal Processing , pages=

work page
[31]

IEEE International Conference on Computer Vision , pages=

Semantic image segmentation via deep parsing network , author=. IEEE International Conference on Computer Vision , pages=

work page
[32]

IEEE conference on computer vision and pattern recognition , pages=

Attention to scale: Scale-aware semantic image segmentation , author=. IEEE conference on computer vision and pattern recognition , pages=

work page
[33]

European Conference on Computer Vision , pages=

Perceptual losses for real-time style transfer and super-resolution , author=. European Conference on Computer Vision , pages=

work page
[34]

, author=

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network. , author=. CVPR , volume=

work page
[35]

Generative Adversarial Text to Image Synthesis

Generative adversarial text to image synthesis , author=. arXiv preprint arXiv:1605.05396 , year=

work page internal anchor Pith review Pith/arXiv arXiv
[36]

arXiv preprint , year=

Stackgan: Text to photo-realistic image synthesis with stacked generative adversarial networks , author=. arXiv preprint , year=

work page
[37]

Computer Vision Workshops (WACVW), 2018 IEEE Winter Applications of , pages=

Facial attributes guided deep sketch-to-photo synthesis , author=. Computer Vision Workshops (WACVW), 2018 IEEE Winter Applications of , pages=. 2018 , organization=

work page 2018
[38]

Conditional Generative Adversarial Nets

Conditional generative adversarial nets , author=. arXiv preprint arXiv:1411.1784 , year=

work page internal anchor Pith review Pith/arXiv arXiv
[39]

IEEE Conference on Computer Vision and Pattern Recognition , year=

High-resolution image inpainting using multi-scale neural patch synthesis , author=. IEEE Conference on Computer Vision and Pattern Recognition , year=

work page
[40]

IEEE Conference on Computer Vision and Pattern Recognition , pages=

Context encoders: Feature learning by inpainting , author=. IEEE Conference on Computer Vision and Pattern Recognition , pages=

work page
[41]

IEEE transactions on image processing , volume=

Face sketch synthesis via sparse representation-based greedy search , author=. IEEE transactions on image processing , volume=

work page
[42]

IEEE transactions on image processing , volume=

Image quality assessment: from error visibility to structural similarity , author=. IEEE transactions on image processing , volume=

work page
[43]

IEEE Conference on Computer Vision and Pattern Recognition , pages=

A non-local algorithm for image denoising , author=. IEEE Conference on Computer Vision and Pattern Recognition , pages=

work page
[44]

arXiv preprint arXiv:2111.11124 , year=

Mesa: A memory-saving training framework for transformers , author=. arXiv preprint arXiv:2111.11124 , year=

work page arXiv
[45]

Arik and Tomas Pfister , title =

Zizhao Zhang and Han Zhang and Long Zhao and Ting Chen and Sercan. Arik and Tomas Pfister , title =. Proceedings of the AAAI Conference on Artificial Intelligence , year =

work page
[46]

IEEE International Conference on Computer Vision , pages=

Deep colorization , author=. IEEE International Conference on Computer Vision , pages=

work page
[47]

IEEE International Conference on Computer Vision , year=

The unreasonable effectiveness of deep features as a perceptual metric , author=. IEEE International Conference on Computer Vision , year=

work page
[48]

IEEE International Conference on Computer Vision , pages=

Least squares generative adversarial networks , author=. IEEE International Conference on Computer Vision , pages=

work page
[49]

ACM Annual Conference on Computer Graphics and Interactive Techniques , pages=

Image quilting for texture synthesis and transfer , author=. ACM Annual Conference on Computer Graphics and Interactive Techniques , pages=

work page
[50]

ACM transactions on graphics , volume=

Removing camera shake from a single photograph , author=. ACM transactions on graphics , volume=

work page
[51]

Advances in neural information processing systems , pages=

Imagenet classification with deep convolutional neural networks , author=. Advances in neural information processing systems , pages=

work page
[52]

IEEE Conference on Computer Vision and Pattern Recognition , year =

He, Kaiming and Zhang, Xiangyu and Ren, Shaoqing and Sun, Jian , title =. IEEE Conference on Computer Vision and Pattern Recognition , year =

work page
[53]

IEEE transactions on pattern analysis and machine intelligence , volume=

3D convolutional neural networks for human action recognition , author=. IEEE transactions on pattern analysis and machine intelligence , volume=

work page
[54]

IEEE Conference on Computer Vision and Pattern Recognition , year =

Alahi, Alexandre and Goel, Kratarth and Ramanathan, Vignesh and Robicquet, Alexandre and Fei-Fei, Li and Savarese, Silvio , title =. IEEE Conference on Computer Vision and Pattern Recognition , year =

work page
[55]

IEEE transactions on medical imaging , volume=

Convolutional neural networks for medical image analysis: Full training or fine tuning? , author=. IEEE transactions on medical imaging , volume=

work page
[56]

IEEE Conference on Computer Vision and Pattern Recognition , pages=

Fine-Tuning Convolutional Neural Networks for Biomedical Image Analysis: Actively and Incrementally , author=. IEEE Conference on Computer Vision and Pattern Recognition , pages=

work page
[57]

IEEE Conference on Computer Vision and Pattern Recognition , year =

Tompson, Jonathan and Goroshin, Ross and Jain, Arjun and LeCun, Yann and Bregler, Christoph , title =. IEEE Conference on Computer Vision and Pattern Recognition , year =

work page
[58]

IEEE Conference on Computer Vision and Pattern Recognition , year =

Zhou, Bolei and Khosla, Aditya and Lapedriza, Agata and Oliva, Aude and Torralba, Antonio , title =. IEEE Conference on Computer Vision and Pattern Recognition , year =

work page
[59]

arXiv preprint arXiv:1703.07195 , year=

Gp-gan: Towards realistic high-resolution image blending , author=. arXiv preprint arXiv:1703.07195 , year=

work page arXiv
[60]

Semantic Segmentation using Adversarial Networks

Semantic segmentation using adversarial networks , author=. arXiv preprint arXiv:1611.08408 , year=

work page internal anchor Pith review Pith/arXiv arXiv
[61]

IEEE Transactions on Image Processing , volume=

Perceptual adversarial networks for image-to-image transformation , author=. IEEE Transactions on Image Processing , volume=

work page
[62]

International Conference on Learning Representations , year=

Adam: A method for stochastic optimization , author=. International Conference on Learning Representations , year=

work page
[63]

Advances in Neural Information Processing Systems , pages=

Toward multimodal image-to-image translation , author=. Advances in Neural Information Processing Systems , pages=

work page
[64]

Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks

Unsupervised representation learning with deep convolutional generative adversarial networks , author=. arXiv preprint arXiv:1511.06434 , year=

work page internal anchor Pith review Pith/arXiv arXiv
[65]

IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=

Face photo-sketch synthesis and recognition , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=. 2009 , publisher=

work page 2009
[66]

Brown and S

M. Brown and S. S\"usstrunk , title =. Computer Vision and Pattern Recognition (CVPR11) , pages =. 2011 , OPTeditor =

work page 2011
[67]

Computer Vision and Pattern Recognition (CVPR), 2011 IEEE Conference on , pages=

Coupled information-theoretic encoding for face photo-sketch recognition , author=. Computer Vision and Pattern Recognition (CVPR), 2011 IEEE Conference on , pages=. 2011 , organization=

work page 2011
[68]

CSGAN: Cyclic-Synthesized Generative Adversarial Networks for Image-to-Image Transformation

CSGAN: Cyclic-Synthesized Generative Adversarial Networks for Image-to-Image Transformation , author=. arXiv preprint arXiv:1901.03554 , year=

work page internal anchor Pith review Pith/arXiv arXiv 1901
[69]

IEEE Transactions on Image Processing , title=

A. IEEE Transactions on Image Processing , title=. 2019 , volume=. doi:10.1109/TIP.2019.2895768 , ISSN=

work page doi:10.1109/tip.2019.2895768 2019
[70]

IEEE Transactions on Image Processing , title=

J. IEEE Transactions on Image Processing , title=. 2019 , volume=. doi:10.1109/TIP.2019.2895460 , ISSN=

work page doi:10.1109/tip.2019.2895460 2019
[71]

IEEE Transactions on Image Processing , title=

Z. IEEE Transactions on Image Processing , title=. 2018 , volume=. doi:10.1109/TIP.2018.2887017 , ISSN=

work page doi:10.1109/tip.2018.2887017 2018
[72]

IEEE Transactions on Image Processing , title=

C. IEEE Transactions on Image Processing , title=. 2018 , volume=. doi:10.1109/TIP.2018.2887029 , ISSN=

work page doi:10.1109/tip.2018.2887029 2018
[73]

IEEE Transactions on Image Processing , title=

J. IEEE Transactions on Image Processing , title=. 2019 , volume=. doi:10.1109/TIP.2018.2869722 , ISSN=

work page doi:10.1109/tip.2018.2869722 2019
[74]

IEEE transactions on neural networks and learning systems , volume=

A Deep Collaborative Framework for Face Photo--Sketch Synthesis , author=. IEEE transactions on neural networks and learning systems , volume=. 2019 , publisher=

work page 2019
[75]

IEEE transactions on neural networks and learning systems , year=

Inverting the generator of a generative adversarial network , author=. IEEE transactions on neural networks and learning systems , year=

work page
[76]

IEEE transactions on neural networks and learning systems , year=

RecSys-DAN: Discriminative Adversarial Networks for Cross-Domain Recommender Systems , author=. IEEE transactions on neural networks and learning systems , year=

work page
[77]

IEEE transactions on neural networks and learning systems , volume=

Face Sketch Synthesis by Multidomain Adversarial Learning , author=. IEEE transactions on neural networks and learning systems , volume=. 2018 , publisher=

work page 2018
[78]

IEEE transactions on neural networks and learning systems , volume=

Multiple representations-based face sketch--photo synthesis , author=. IEEE transactions on neural networks and learning systems , volume=. 2015 , publisher=

work page 2015
[79]

IEEE transactions on neural networks and learning systems , volume=

Transductive face sketch-photo synthesis , author=. IEEE transactions on neural networks and learning systems , volume=. 2013 , publisher=

work page 2013
[80]

IEEE transactions on neural networks and learning systems , volume=

Robust kernel representation with statistical local features for face recognition , author=. IEEE transactions on neural networks and learning systems , volume=. 2013 , publisher=

work page 2013

Showing first 80 references.