Pith. sign in

REVIEW 3 major objections 5 minor 43 references

SyncMapV2: Robust and Adaptive Unsupervised Segmentation

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

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

Pith's one-line read SyncMapV2 is claimed to be the first unsupervised segmentation method with state-of-the-art robustness, losing only 0.01% mIoU under digital corruption while SOTA methods lose 23.8%.

desk verdict Interesting method, but the SOTA robustness claim is undermined by an asymmetric evaluation (OIS for SyncMapV2, ODS for baselines) and abstract numbers that do not match the tables. read the letter →

arxiv 2506.16297 v3 pith:NIIYQZGL submitted 2025-06-19 cs.CV cs.AIcs.LG

Heng Zhang , Zikang Wan , Danilo Vasconcellos Vargas This is my paper · ORCID
classification cs.CVcs.AIcs.LG
keywords unsupervisedimagesegmentationself-organizingdynamicalsystemsreservoircomputingechostatenetworkdynamictimewarpingrobustnesstocorruptiononlineadaptabilityno-traininglearning
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

SyncMapV2 aims to show that unsupervised image segmentation can be made robust to image corruption and continuously adaptive without any training, loss functions, or re-initialization. The paper reports that on corrupted versions of the BSD500 dataset, its method loses only 7.0–7.5% in mIoU across noise, blur, weather, and digital contrast, while the leading trained baseline DFC loses 23.8–37.7%. On clean data SyncMapV2 trails the state of the art by 12.3%, so the claim is a trade: robust and adaptive features come from self-organization rather than peak accuracy on a benchmark. If correct, this opens a training-free route to perception systems that keep working as their environment changes.

What carries the argument

The central mechanism is the SyncMap dynamical system—a set of attractor-repeller update equations that pull co-active state variables together in a map space—driven by a per-step binary input vector built from patch proximity (the eight spatial neighbours of a randomly chosen reference patch) and patch similarity (the top patches ranked by Dynamic Time Warping between the reservoir responses of two patches). The reservoir is an untrained echo state network that converts each patch's RGB content into a temporal sequence of neuron states; DTW then scores structural similarity. Five stabilising modifications—adaptive learning rate, symmetrical activations, zero-mean unit-variance space normalization, a moving average over map positions, and a leaking rate—make the dynamics converge to stable clusters that are then read out by hierarchical clustering.

What would settle it

Run SyncMapV2 on the BSD500 test set under all 15 corruption types of the standard common-corruption benchmark at all five severities (the paper tests only four corruptions); if any corruption type—such as impulse noise, JPEG compression, or frost—produces an mIoU drop greater than 15 percentage points relative to clean, the paper's claim of state-of-the-art, near-universal robustness would be contradicted.

Watch

Extended reading notes

Core claim

The paper's central discovery is that a self-organizing dynamical system fed with a sequence of image patches—selected by spatial proximity and by similarity measured through DTW on the responses of an untrained echo state network—produces segmentation that is nearly invariant to common image corruptions. SyncMapV2 is claimed to be the first unsupervised segmentation method with state-of-the-art robustness: its mIoU drops only 0.01% under digital contrast corruption versus a 23.8% drop for DFC, and it shows similar margins under noise, weather, and blur. The system also adapts online: when the input image changes during a single run of the dynamics, accuracy does not degrade, whereas loss-based methods collapse to near chance without re-initialization. The authors attribute this to the absence of loss-driven feature learning: correlations are extracted from structural, population-level responses of a random reservoir rather than from trained weights.

Load-bearing premise

The results rest on the untested premise that DTW distances computed on the responses of a randomly initialised reservoir reflect true segment boundaries and stay stable when images are corrupted; the paper provides no theoretical or error analysis showing when this similarity is corruption-invariant.

Editorial extensions

If this is right

  • Unsupervised segmentation can be performed without any training, supervision, or loss function, with robustness exceeding current state-of-the-art loss-trained models under the four tested corruption families.
  • Robustness and clean-data accuracy trade off: the method is 12.3% below SOTA on clean data yet far better on corrupted data, so progress on robustness may require valuing stable features over benchmark peaks.
  • A single continuous run of the dynamics can segment an arbitrary sequence of images without re-initialization, which is the first demonstrated case of online adaptability for unsupervised segmentation.
  • Because the pipeline uses only untrained random-network responses and DTW, the same mechanism may extend to other tasks that require grouping correlated temporal or spatial events.
  • Loss function-based segmentation methods become effectively non-adaptive after their loss converges, whereas self-organizing dynamics re-adapt by changing attractor positions when the input changes.

Reading between the lines

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

  • If the reservoir-similarity mechanism is the true source of robustness, the method may generalize to the full corruption benchmark and to unseen distortions; this can be tested directly by extending the corruption suite.
  • The trade-off between clean accuracy and robustness suggests a promising hybrid: use trained features for clean images but fall back to a self-organizing readout when corruption is detected.
  • The same DTW-reservoir similarity could be applied to video or point-cloud data, where temporal structure is natural, extending the method beyond 2D images.
  • The lack of a proof of DTW stability implies the robustness may be dataset- or corruption-specific; a theoretical characterization of reservoir similarity under perturbation would turn an empirical result into a mechanism.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes SyncMapV2, a no-training unsupervised image segmentation method. It encodes image patches through an untrained echo state network, computes DTW-based similarities between the resulting reservoir responses, and feeds binary activation sequences into self-organizing attractor/repeller dynamics in a map space; hierarchical clustering produces the final segmentation. Experiments on VOC2012 and BSD500 report state-of-the-art clean mIoU on VOC2012 and robustness results on BSD500 under four corruption types, plus online adaptability without re-initialization. The abstract claims near-zero mIoU degradation (0.01%) under digital corruption, a 23.8% drop for SOTA methods, and near-zero degradation in adaptability tests.

Significance. If the headline result held, a completely untrained dynamical system that segments images and remains stable under corruption would be a noteworthy contribution, both as an unsupervised segmentation method and as evidence for the representational-space approach. The paper provides algorithmic equations, ablations isolating each component, and statistical tests, which are useful. However, the central robustness claim is not yet trustworthy because the evaluation protocol is asymmetric between SyncMapV2 and DFC on the critical OIS/ODS cluster-count choice, and the abstract percentages do not reproduce from Table 2. These issues must be resolved before the robustness and adaptability claims can be assessed.

major comments (3)
  1. [§5.3, Table 2; §5.2, Table 1] The robustness comparison between SyncMapV2 and DFC is confounded by an asymmetric cluster-count protocol. In Table 1, only SyncMap and SyncMapV2 are marked with '*' for OIS; DFC's BSD500 value 0.3739 is unstarred, i.e., ODS. In Table 2, DFC's clean value is 0.37 (matching its unstarred ODS value 0.3739) and SyncMapV2's clean value is 0.33 (matching its starred OIS value 0.3279). Section 5.2 states that OIS is used for SyncMapV2 because of variability in BSD500 segment numbers, but the paper does not state that DFC receives the same treatment. OIS selects the optimal number of clusters per image using ground truth, which can inflate measured accuracy and reduce measured degradation under corruption. This is load-bearing for the central claim of a 0.01% versus 23.8% drop. Please report both ODS and OIS for both methods in Table 2, or otherwise use a fixed cluster-count protocol for all methods.
  2. [Abstract; §5.3, Table 2] The percentage drops reported in the abstract do not match Table 2. From Table 2, DFC's digital mean is 0.26 versus clean 0.37, which is a 29.7% drop, not the claimed 23.8%. SyncMapV2's noise mean is 0.30 versus 0.33, which is a 9.1% drop, not the claimed 7.3%. With the unrounded OIS clean value 0.3279, SyncMapV2's digital mean is about 0.3267, giving a drop of roughly 0.37%, not 0.01%. The weather, blur, and Figure 5 percentages also differ from the table. Recompute all reported percentages from the underlying per-image scores and ensure that the abstract, Section 5.3, Figure 5, and Table 2 are mutually consistent.
  3. [§5.3, Table 2; §4.2] The robustness claim rests on only four corruption types (Gaussian noise, zoom blur, snow weather, digital contrast) on one dataset, and Table 2 reports only mean mIoU without per-image standard errors or confidence intervals. Table 5 gives p-values for aggregate comparisons, but not the variability that would let a reader judge whether the near-zero degradation claim is stable. Because Section 4.2 offers no analysis or guarantee that DTW distances between untrained reservoir responses are corruption-invariant, the generalization of the robustness result beyond these four corruptions is undetermined. Reporting per-image variance and the full corruption benchmark from [40], or a clear selection rationale, is needed to support the central robustness claim.
minor comments (5)
  1. [Title] The title is typeset as 'SYNC MAPV2' rather than 'SyncMapV2' in the header; please correct the capitalization.
  2. [Appendix B.1 and §3.5] The leaking rate is denoted α in the ESN equations (with a stated value of 0.5 in B.1) and β in Section 3.5 (with a stated value of 0.1); the relationship between these two rates should be clarified.
  3. [Figure 2] Figure 2 is dense and the fonts for the patch-selection and DTW panels are very small; increasing the size and separating the preprocessing and learning stages would improve readability.
  4. [Algorithm 2] The pseudocode in Appendix B.4.3 uses informal phrases such as 'smaller than' and 'greater than'; it should be replaced with standard comparison operators.
  5. [Table 3] The lower-bound row in Table 3 lists the same value 0.1642 three times, and DFC's w/o erl.stp value 0.1657 is very close to that bound; the table would be clearer if the bound were a single column or clearly annotated.

Circularity Check

1 steps flagged · score 6.0 of 10

Robustness headline is partially an artifact of asymmetric OIS/ODS evaluation: SyncMapV2's cluster count is oracle-fitted per image while DFC's is fixed.

  1. fitted input called prediction [Section 5.1.3 (Evaluation Metrics), Table 1 caption, Section 5.2, and Section 5.3 / Table 2]
    "We report two performance metrics: Optimal Dataset Scale (ODS), which evaluates a fixed nclu across the dataset, and Optimal Image Scale (OIS), which is widely used to select the optimal nclu for each image [21, 38, 39]. ... The variability in segment numbers within BSD500 makes OIS metric a more effective measure for visualizing the performance of SyncMapV2. ... Table 1 caption: The '*' denotes OIS results."

    Table 1 gives SyncMapV2 an OIS BSD500 score of 0.3279 and DFC an unstarred ODS score of 0.3739; Table 2 uses the same values as the Clean row (0.33 vs 0.37) before computing corruption drops. OIS means that SyncMapV2's reported mIoU is, by definition, the maximum over nclu chosen per image against ground truth, while DFC is held to one dataset-wide cluster setting. The advertised claim that SyncMapV2 loses only 0.01% under digital corruption versus DFC's 23.8% therefore compares an oracle-optimized cluster count for SyncMapV2 with a fixed one for DFC; part of the apparent robustness is inserted by the evaluation metric rather than produced by the no-training dynamics. The per-image cluster count is a ground-truth-fitted parameter being reported as a model property.

full rationale

The core machinery of SyncMapV2 is a self-organizing dynamical system evaluated on external benchmarks, so most of the paper is empirical rather than a derivation that reduces to its inputs. The self-citations to the original SyncMap [12] and symmetrical activation [31] are acknowledged and are not used as uniqueness theorems to forbid alternatives; the ablation study tests the components. However, the central robustness comparison is partially circular in the sense of a fitted parameter presented as a prediction: the reported performance of SyncMapV2 on BSD500 and on the corrupted robustness test is OIS, i.e., per image the number of clusters is chosen to maximize mIoU against ground truth, whereas DFC's numbers are ODS, a fixed cluster count over the dataset. Consequently the headline '0.01% vs 23.8% drop' is not a like-for-like measurement of the dynamics' stability and is partly forced by the evaluation protocol. In addition, the abstract's percentages are internally inconsistent with Table 2, for example, DFC's digital mean of 0.26 versus clean 0.37 is about a 29.7% drop rather than the claimed 23.8%, which further weakens the quantitative claim. These are evaluation and reporting defects rather than a self-citation chain; the method's output on individual images is still genuinely produced by the dynamical equations, so the circularity is partial, not total.

Assumptions & free parameters 9 free parameters · 4 assumptions · 0 invented entities

The central claims rest on a long chain of hand-set hyperparameters and on assumptions that untrained reservoir responses and DTW similarities remain stable under corruption; no code is provided. These do not amount to circular derivations, but they make the empirical claims less robust than the prose suggests.

free parameters (9)
  • Number of clusters nclu per image via OIS = 2 to 20, selected per image to maximize mIoU
    OIS is an oracle selector; the benchmark and robustness numbers labeled with asterisks use it, and the tables do not state whether DFC also receives the same oracle treatment.
  • SyncMapV2 space dimension k = 15
    Chosen by hand in Section 5.1.3; controls convergence speed and cluster geometry.
  • Adaptive positive learning rate parameters = alpha+ = meanDistance/sqrt(k), floor 0.05, smoothed by adaptive_LR_WH
    Algorithm 2 introduces an exponential moving average and thresholds; these are hand-set and influence convergence.
  • Negative learning rate parameters = alpha- base 1, amplifier 0.01*input_size + 2, cap 1.5
    Algorithm 2 in Appendix B.4; constants chosen by hand.
  • Leaking rate for SyncMapV2 dynamics = 0.1
    Section 3.5; hand-set.
  • Moving average window = 2000 steps
    Section 3.4; hand-set.
  • Patch grid and image resize = 288x288 image, 48x48 patches, top-9 proximity and similarity lists
    Section 5.1.3; hand-set to balance speed and accuracy.
  • ESN configuration = 512 units, sparsity 0.9, spectral radius 1.1, input scaling 1.0, reservoir leaking rate 0.5, K=3 repetitions
    Appendix B.1; the unsupervised pipeline depends on these reservoir choices.
  • Sequence length tau = 200,000 steps; 400,000 for 2tau
    Section 5.1.3; hand-set, with 2tau used in the adaptability test.
assumptions (4)
  • domain assumption The same untrained ESN can map every image patch into comparable reservoir responses whose DTW distances reflect perceptual similarity.
    Section 4.2 defines the similarity matrix this way; no proof is given that the representation is invariant to semantic class or corruption.
  • domain assumption Echo state property holds for the chosen reservoir parameters, including spectral radius 1.1.
    Appendix B.1 sets spectral radius 1.1, above the conventional echo state condition; the network is assumed stable enough for comparisons.
  • ad hoc to paper Proximity and similarity lists are sufficient for segmenting arbitrary natural images.
    Section 4 bases the input construction on Gestalt principles; there is no guarantee for objects split by textures or large uniform regions.
  • standard math Hierarchical clustering of the final map yields the ground-truth segments.
    Sections 2.3 and 4.1 use hierarchical clustering as readout; this is a standard tool, but its output depends on the chosen cluster count.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SyncMapV2: Robust and Adaptive Unsupervised Segmentation." pith.science (2026). https://pith.science/paper/NIIYQZGL

@misc{pith2026250616297,
  author       = {Pith},
  title        = {Pith review of: SyncMapV2: Robust and Adaptive Unsupervised Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NIIYQZGL}},
  note         = {Machine review of arXiv:2506.16297}
}
read the original abstract

Human vision excels at segmenting visual cues without the need for explicit training, and it remains remarkably robust even as noise severity increases. In contrast, existing AI algorithms struggle to maintain accuracy under similar conditions. Here, we present SyncMapV2, the first to solve unsupervised segmentation with state-of-the-art robustness. SyncMapV2 exhibits a minimal drop in mIoU, only 0.01%, under digital corruption, compared to a 23.8% drop observed in SOTA methods. This superior performance extends across various types of corruption: noise (7.3% vs. 37.7%), weather (7.5% vs. 33.8%), and blur (7.0% vs. 29.5%). Notably, SyncMapV2 accomplishes this without any robust training, supervision, or loss functions. It is based on a learning paradigm that uses self-organizing dynamical equations combined with concepts from random networks. Moreover, unlike conventional methods that require re-initialization for each new input, SyncMapV2 adapts online, mimicking the continuous adaptability of human vision. Thus, we go beyond the accurate and robust results, and present the first algorithm that can do all the above online, adapting to input rather than re-initializing. In adaptability tests, SyncMapV2 demonstrates near-zero performance degradation, which motivates and fosters a new generation of robust and adaptive intelligence in the near future.

Figures

Figures reproduced from arXiv: 2506.16297 by the authors.

Figure 1
Figure 1. Illustration of SyncMapV2’s dynamics. Note that the random patches are selected apart from positive patches. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of the preprocessing and learning pipeline for unsupervised image segmentation using SyncMapV2. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Comparison of segmentation on standard tests. Different segments are shown in different colors. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: SyncMapV2 performs robustly across four types of algorithmically generated corruptions from noise, blur, [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: SyncMapV2 outperforms the SOTA method (DFC) across all corrupted conditions (top). Vertically, DFC [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: We propose the first adaptability test pipeline, where model initialization occurs only once. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Sample results of adaptability tests. SyncMapV2 consistently adapts to each input, while DFC fails to adapt [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: Adaptability plots over time for self-organization-based SyncMapV2 (green, top) and loss function-based [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Simplified structure of Echo State Network (ESN). [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 31 canonical work pages

  1. [40]

    Benchmarking neural network robustness to common corruptions and perturbations

    Dan Hendrycks and Thomas Dietterich. Benchmarking neural network robustness to common corruptions and perturbations. Proceedings of the International Conference on Learning Representations, 2019. 18

    work page 2019

  2. [1]

    Why do deep convolutional networks generalize so poorly to small image transformations? Journal of Machine Learning Research, 20(184):1–25, 2019

    Aharon Azulay and Yair Weiss. Why do deep convolutional networks generalize so poorly to small image transformations? Journal of Machine Learning Research, 20(184):1–25, 2019. 16 Table 6: Paired Two-Sided T-Test of SyncMapV2 P-value. Avg: Average mIoU of severity 1, 3, and 5. We compare the average performance of each corruption type with performance on c...

    work page 2019

  3. [2]

    Do cifar-10 classifiers generalize to cifar-10? arXiv preprint arXiv:1806.00451, 2018

    Benjamin Recht, Rebecca Roelofs, Ludwig Schmidt, and Vaishaal Shankar. Do cifar-10 classifiers generalize to cifar-10? arXiv preprint arXiv:1806.00451, 2018

    arXiv 2018

  4. [3]

    Intriguing properties of neural networks

    Szegedy Cea. Intriguing properties of neural networks. In ICLR. Citeseer, 2014

    work page 2014

  5. [4]

    Explaining and harnessing adversarial examples

    Ian J Goodfellow, Jonathon Shlens, and Christian Szegedy. Explaining and harnessing adversarial examples. arXiv preprint arXiv:1412.6572, 2014

    arXiv 2014

  6. [5]

    Deep neural networks are easily fooled: High confidence predictions for unrecognizable images

    Anh Nguyen, Jason Yosinski, and Jeff Clune. Deep neural networks are easily fooled: High confidence predictions for unrecognizable images. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 427–436, 2015

    work page 2015

  7. [6]

    Adversarial robustness assessment: Why in evaluation both L0 and L∞ attacks are necessary

    Shashank Kotyan and Danilo Vasconcellos Vargas. Adversarial robustness assessment: Why in evaluation both L0 and L∞ attacks are necessary. Plos one, 17(4):e0265723, 2022

    work page 2022

  8. [7]

    Thermometer encoding: One hot way to resist adversarial examples

    Jacob Buckman, Aurko Roy, Colin Raffel, and Ian Goodfellow. Thermometer encoding: One hot way to resist adversarial examples. In International conference on learning representations, 2018

    work page 2018

Show all 43 references
  1. [8]

    Adversarial risk and the dangers of evaluating against weak attacks

    Jonathan Uesato, Brendan O’donoghue, Pushmeet Kohli, and Aaron Oord. Adversarial risk and the dangers of evaluating against weak attacks. In International Conference on Machine Learning, pages 5025–5034. PMLR, 2018

    2018

  2. [9]

    Towards deep learning models resistant to adversarial attacks

    Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, and Adrian Vladu. Towards deep learning models resistant to adversarial attacks. arXiv preprint arXiv:1706.06083, 2017

    2017 arXiv

  3. [10]

    Feature squeezing: Detecting adversarial examples in deep neural networks

    Weilin Xu, David Evans, and Yanjun Qi. Feature squeezing: Detecting adversarial examples in deep neural networks. arXiv preprint arXiv:1704.01155, 2017

    2017 arXiv

  4. [11]

    Adversarial examples are not easily detected: Bypassing ten detection methods

    Nicholas Carlini and David Wagner. Adversarial examples are not easily detected: Bypassing ten detection methods. In Proceedings of the 10th ACM workshop on artificial intelligence and security, pages 3–14, 2017

    2017

  5. [12]

    Continual general chunking problem and syncmap

    Danilo Vasconcellos Vargas and Toshitake Asabuki. Continual general chunking problem and syncmap. Proceed- ings of the AAAI Conference on Artificial Intelligence, 35(11):10006–10014, 5 2021

    2021

  6. [13]

    Unsupervised universal image segmentation

    Dantong Niu, Xudong Wang, Xinyang Han, Long Lian, Roei Herzig, and Trevor Darrell. Unsupervised universal image segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 22744–22754, 2024

    2024

  7. [14]

    Unsupervised semantic segmentation by distilling feature correspondences

    Mark Hamilton, Zhoutong Zhang, Bharath Hariharan, Noah Snavely, and William T Freeman. Unsupervised semantic segmentation by distilling feature correspondences. arXiv preprint arXiv:2203.08414, 2022

    2022 arXiv

  8. [15]

    Comaniciu and P

    D. Comaniciu and P. Meer. Mean shift: a robust approach toward feature space analysis. IEEE Transactions on Pattern Analysis and Machine Intelligence, 24(5):603–619, 2002

    2002

  9. [16]

    Efficient graph-based image segmentation

    Pedro F Felzenszwalb and Daniel P Huttenlocher. Efficient graph-based image segmentation. International journal of computer vision, 59:167–181, 2004

    2004

  10. [17]

    W-net: A deep model for fully unsupervised image segmentation

    Xide Xia and Brian Kulis. W-net: A deep model for fully unsupervised image segmentation. arXiv preprint arXiv:1711.08506, 2017. 17

    2017 arXiv

  11. [18]

    Unsupervised image segmentation by backpropagation

    Asako Kanezaki. Unsupervised image segmentation by backpropagation. In 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1543–1547, 2018

    2018

  12. [19]

    Invariant information clustering for unsupervised image classification and segmentation

    Xu Ji, Andrea Vedaldi, and Joao Henriques. Invariant information clustering for unsupervised image classification and segmentation. In 2019 IEEE/CVF International Conference on Computer Vision (ICCV), pages 9864–9873, 2019

    2019

  13. [20]

    Unsupervised learning of image segmentation based on differentiable feature clustering

    Wonjik Kim, Asako Kanezaki, and Masayuki Tanaka. Unsupervised learning of image segmentation based on differentiable feature clustering. IEEE Transactions on Image Processing, 29:8055–8068, 2020

    2020

  14. [21]

    Pixel-level clustering network for unsupervised image segmentation

    Cuong Manh Hoang and Byeongkeun Kang. Pixel-level clustering network for unsupervised image segmentation. Engineering Applications of Artificial Intelligence, 127:107327, 2024

    2024

  15. [22]

    Two views on the cognitive brain

    David L Barack and John W Krakauer. Two views on the cognitive brain. Nature Reviews Neuroscience, 22(6):359–371, 2021

    2021

  16. [23]

    Accurate estimation of neural population dynamics without spike sorting

    Eric M Trautmann, Sergey D Stavisky, Subhaneil Lahiri, Katherine C Ames, Matthew T Kaufman, Daniel J O’Shea, Saurabh Vyas, Xulu Sun, Stephen I Ryu, Surya Ganguli, et al. Accurate estimation of neural population dynamics without spike sorting. Neuron, 103(2):292–308, 2019

    2019

  17. [24]

    The importance of mixed selectivity in complex cognitive tasks

    Mattia Rigotti, Omri Barak, Melissa R Warden, Xiao-Jing Wang, Nathaniel D Daw, Earl K Miller, and Stefano Fusi. The importance of mixed selectivity in complex cognitive tasks. Nature, 497(7451):585–590, 2013

    2013

  18. [25]

    A survey on reservoir computing and its interdisciplinary applica- tions beyond traditional machine learning

    Heng Zhang and Danilo Vasconcellos Vargas. A survey on reservoir computing and its interdisciplinary applica- tions beyond traditional machine learning. IEEE Access, 2023

    2023

  19. [26]

    Smooseg: smoothness prior for unsupervised semantic segmentation

    Mengcheng Lan, Xinjiang Wang, Yiping Ke, Jiaxing Xu, Litong Feng, and Wayne Zhang. Smooseg: smoothness prior for unsupervised semantic segmentation. Advances in Neural Information Processing Systems, 36, 2024

    2024

  20. [27]

    Diffuse attend and segment: Unsupervised zero-shot segmentation using stable diffusion

    Junjiao Tian, Lavisha Aggarwal, Andrea Colaco, Zsolt Kira, and Mar Gonzalez-Franco. Diffuse attend and segment: Unsupervised zero-shot segmentation using stable diffusion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3554–3563, 2024

    2024

  21. [28]

    Unsupervised semantic segmentation through depth-guided feature correlation and sampling

    Leon Sick, Dominik Engel, Pedro Hermosilla, and Timo Ropinski. Unsupervised semantic segmentation through depth-guided feature correlation and sampling. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3637–3646, 2024

    2024

  22. [29]

    Dbscan revisited, revisited: why and how you should (still) use dbscan

    Erich Schubert, Jörg Sander, Martin Ester, Hans Peter Kriegel, and Xiaowei Xu. Dbscan revisited, revisited: why and how you should (still) use dbscan. ACM Transactions on Database Systems (TODS), 42(3):1–21, 2017

    2017

  23. [30]

    Algorithms for hierarchical clustering: an overview

    Fionn Murtagh and Pedro Contreras. Algorithms for hierarchical clustering: an overview. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 2(1):86–97, 2012

    2012

  24. [31]

    Symmetrical syncmap for imbalanced general chunking problems

    Heng Zhang and Danilo Vasconcellos Vargas. Symmetrical syncmap for imbalanced general chunking problems. Physica D: Nonlinear Phenomena, 456:133923, 2023

    2023

  25. [32]

    Gestalt psychology

    Max Wertheimer. Gestalt psychology. Source Book of Gestalt Psychology. New York: Harcourt, Brace and Co, 1938

    1938

  26. [33]

    Principles of neural science, volume 4

    Eric R Kandel, James H Schwartz, Thomas M Jessell, Steven Siegelbaum, A James Hudspeth, Sarah Mack, et al. Principles of neural science, volume 4. McGraw-hill New York, 2000

    2000

  27. [34]

    echo state

    Herbert Jaeger. The “echo state” approach to analysing and training recurrent neural networks-with an erratum note. Bonn, Germany: German National Research Center for Information Technology GMD Technical Report, 148(34):13, 2001

    2001

  28. [35]

    Computing and visualizing dynamic time warping alignments in r: the dtw package

    Toni Giorgino. Computing and visualizing dynamic time warping alignments in r: the dtw package. Journal of statistical Software, 31:1–24, 2009

    2009

  29. [36]

    The pascal visual object classes challenge: A retrospective

    Mark Everingham, SM Ali Eslami, Luc Van Gool, Christopher KI Williams, John Winn, and Andrew Zisserman. The pascal visual object classes challenge: A retrospective. International journal of computer vision, 111:98–136, 2015

    2015

  30. [37]

    Contour detection and hierarchical image segmentation

    Pablo Arbelaez, Michael Maire, Charless Fowlkes, and Jitendra Malik. Contour detection and hierarchical image segmentation. IEEE transactions on pattern analysis and machine intelligence, 33(5):898–916, 2010

    2010

  31. [38]

    Dic: deep image clustering for unsupervised image segmentation

    Lei Zhou and Weiyufeng Wei. Dic: deep image clustering for unsupervised image segmentation. Ieee Access, 8:34481–34491, 2020

    2020

  32. [39]

    Segmentation using superpixels: A bipartite graph partitioning approach

    Zhenguo Li, Xiao-Ming Wu, and Shih-Fu Chang. Segmentation using superpixels: A bipartite graph partitioning approach. In 2012 IEEE conference on computer vision and pattern recognition, pages 789–796. IEEE, 2012

    2012

  33. [41]

    Dynaseg: A deep dynamic fusion method for unsuper- vised image segmentation incorporating feature similarity and spatial continuity

    Boujemaa Guermazi, Riadh Ksantini, and Naimul Khan. Dynaseg: A deep dynamic fusion method for unsuper- vised image segmentation incorporating feature similarity and spatial continuity. Image and Vision Computing, page 105206, 2024

    2024

  34. [42]

    A practical guide to applying echo state networks

    Mantas Lukoševiˇcius. A practical guide to applying echo state networks. In Neural networks: Tricks of the trade, pages 659–686. Springer, 2012

    2012

  35. [43]

    Reservoir computing approaches to recurrent neural network training

    Mantas Lukoševiˇcius and Herbert Jaeger. Reservoir computing approaches to recurrent neural network training. Computer Science Review, 3(3):127–149, 2009. 19

    2009

Pith tools

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