pith. sign in

arxiv: 2602.19423 · v4 · pith:PYFXKIN4new · submitted 2026-02-23 · 💻 cs.CV

Prefer-DAS: Learning from Local Preferences and Sparse Prompts for Domain Adaptive Segmentation of Electron Microscopy

Jiabao Chen , Shan Xiong , Jialin Peng This is my paper

Pith reviewed 2026-05-21 13:11 UTC · model grok-4.3

classification 💻 cs.CV
keywords domain adaptive segmentationelectron microscopysparse promptspreference optimizationweakly supervised learninginteractive segmentationself-trainingcontrastive learning
0
0 comments X

The pith

Prefer-DAS adapts electron microscopy segmentation to new domains using sparse point prompts and local human preferences instead of dense labels.

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

The paper develops Prefer-DAS to solve domain adaptive segmentation of intracellular structures in large-scale electron microscopy images when full target-domain annotations are unavailable. It treats sparse points and local human preferences as weak supervision signals and builds a promptable multitask model that combines self-training with prompt-guided contrastive learning. Local direct Preference Optimization aligns the model to spatially varying feedback while Unsupervised Preference Optimization fills gaps when feedback is missing. Experiments across four tasks show the approach works in both automatic and interactive modes and reaches performance close to fully supervised models.

Core claim

Prefer-DAS is a promptable multitask model that integrates self-training and prompt-guided contrastive learning to perform domain adaptive segmentation. It supports full, partial, or no point prompts at both training and inference time, enabling interactive use. Local direct Preference Optimization provides plug-and-play alignment with spatially varying human feedback, and Unsupervised Preference Optimization leverages self-learned preferences to handle missing feedback, allowing the same model to operate in weakly-supervised or unsupervised regimes depending on available inputs.

What carries the argument

Prefer-DAS, a promptable multitask model that performs sparse promptable learning and local preference alignment via Local direct Preference Optimization and Unsupervised Preference Optimization.

If this is right

  • The model supports interactive segmentation by accepting full, partial, or zero point prompts at inference.
  • It delivers performance close to or exceeding supervised baselines on four challenging domain-shift tasks.
  • It outperforms SAM-like methods as well as prior unsupervised and weakly-supervised DAS approaches in both automatic and interactive settings.
  • The same architecture switches between weakly-supervised and unsupervised modes based on the presence of points and preferences.

Where Pith is reading between the lines

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

  • The local-preference alignment strategy could transfer to other volumetric imaging modalities that suffer domain shifts, such as cryo-electron tomography or light-sheet microscopy.
  • Automating preference collection through lightweight active-learning loops would further lower the human cost beyond the current sparse setup.
  • Extending the promptable backbone to handle 3D volumetric inputs directly would address more complex intracellular structures without slice-by-slice processing.

Load-bearing premise

Local human preferences and sparse points in the target domain supply sufficient unbiased and spatially representative supervision that LPO and UPO can align without new biases or prohibitive effort.

What would settle it

Running the model on a fifth unseen EM domain and finding Dice scores more than 5 percent below a fully supervised baseline while annotation time exceeds the reported sparse-prompt cost would falsify the central performance claim.

Figures

Figures reproduced from arXiv: 2602.19423 by Jiabao Chen, Jialin Peng, Shan Xiong.

Figure 1
Figure 1. Figure 1: A visualization of the concept of local human preferences and [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overall framework of the proposed Prefer-DAS model for weakly-supervised domain adaptive segmentation. Our framework comprises two stages. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Visual comparison of segmentation results in both automatic and interactive modes. Despite the Med-SAM Adapter+ and WeSAM+ being fine-tuned [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison of different strategies for preference learning. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Bar plot of the consistency between human preferences on the local [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Effectiveness of preference learning in correcting defective and [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
read the original abstract

Domain adaptive segmentation (DAS) is a promising paradigm for delineating intracellular structures from various large-scale electron microscopy (EM) without incurring extensive annotated data in each domain. However, the prevalent unsupervised domain adaptation (UDA) strategies often demonstrate limited and biased performance, which hinders their practical applications. In this study, we explore sparse points and local human preferences as weak labels in the target domain, thereby presenting a more realistic yet annotation-efficient setting. Specifically, we develop Prefer-DAS, which pioneers sparse promptable learning and local preference alignment. The Prefer-DAS is a promptable multitask model that integrates self-training and prompt-guided contrastive learning. Unlike SAM-like methods, the Prefer-DAS allows for the use of full, partial, and even no point prompts during both training and inference stages and thus enables interactive segmentation. Instead of using image-level human preference alignment for segmentation, we introduce Local direct Preference Optimization (LPO), plug-and-play solutions for alignment with spatially varying human feedback. To address potential missing feedback, we also introduce Unsupervised Preference Optimization (UPO), which leverages self-learned preferences. As a result, the Prefer-DAS model can effectively perform both weakly-supervised and unsupervised DAS, depending on the availability of points and human preferences. Comprehensive experiments on four challenging DAS tasks demonstrate that our model outperforms SAM-like methods as well as unsupervised and weakly-supervised DAS methods in both automatic and interactive segmentation modes, highlighting strong generalizability and flexibility. Additionally, the performance of our model is very close to or even exceeds that of supervised models.

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.

Referee Report

2 major / 2 minor

Summary. The paper introduces Prefer-DAS, a promptable multitask model for domain adaptive segmentation (DAS) in electron microscopy that combines self-training with prompt-guided contrastive learning. It proposes Local direct Preference Optimization (LPO) to align with spatially varying human feedback from sparse points and local preferences in the target domain, plus Unsupervised Preference Optimization (UPO) to handle missing feedback. The model supports automatic and interactive segmentation with full, partial, or no prompts. Comprehensive experiments on four DAS tasks claim outperformance over SAM-like methods, UDA, and weakly-supervised approaches, with performance very close to or exceeding supervised models.

Significance. If the supervised baseline comparisons use matched architectures, capacity, and protocols, the work could meaningfully advance annotation-efficient DAS for EM by demonstrating that sparse local preferences suffice for near-supervised accuracy while enabling interactive use. The LPO/UPO components offer a concrete mechanism for incorporating human feedback without full annotations, which would be a useful addition to the self-training and contrastive learning literature if the empirical gains are robust.

major comments (2)
  1. [Abstract / Results] Abstract and experimental results: The headline claim that performance is 'very close to or even exceeds that of supervised models' on four DAS tasks is central to the generalizability argument. This comparison is only interpretable as evidence of effective adaptation if the supervised baselines employ the identical backbone, model capacity, training protocol, and data splits as Prefer-DAS; otherwise the reported proximity to supervised performance may reflect implementation differences rather than the efficacy of LPO/UPO or sparse prompts.
  2. [Methods] Methods section on LPO and UPO: The integration of local human preferences via LPO and self-learned preferences via UPO assumes these signals are spatially representative and unbiased. The manuscript should provide the explicit loss formulations or algorithmic steps (e.g., how preference pairs are constructed from sparse points and how they are aligned without introducing new spatial bias) to allow verification that the alignment steps do not undermine the comparison to full-supervision upper bounds.
minor comments (2)
  1. [Abstract] The abstract would be clearer if it briefly named the four specific DAS tasks or EM datasets used, rather than referring only to 'four challenging DAS tasks.'
  2. [Methods] Notation for LPO and UPO should be introduced with consistent symbols when first defined to aid readability across the methods and experiments.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major comment point by point below, providing clarifications and indicating revisions to the manuscript where appropriate.

read point-by-point responses

  1. Referee: [Abstract / Results] Abstract and experimental results: The headline claim that performance is 'very close to or even exceeds that of supervised models' on four DAS tasks is central to the generalizability argument. This comparison is only interpretable as evidence of effective adaptation if the supervised baselines employ the identical backbone, model capacity, training protocol, and data splits as Prefer-DAS; otherwise the reported proximity to supervised performance may reflect implementation differences rather than the efficacy of LPO/UPO or sparse prompts.

    Authors: We agree that fair interpretation of the headline claim requires matched supervised baselines. In the experiments reported in the manuscript, the supervised models were trained using the identical backbone architecture, model capacity, training protocol, and data splits as Prefer-DAS. To eliminate any ambiguity, we have revised the abstract and the results section to explicitly state this equivalence and have added a dedicated paragraph in the experimental setup clarifying the implementation details for all baselines. revision: yes

  2. Referee: [Methods] Methods section on LPO and UPO: The integration of local human preferences via LPO and self-learned preferences via UPO assumes these signals are spatially representative and unbiased. The manuscript should provide the explicit loss formulations or algorithmic steps (e.g., how preference pairs are constructed from sparse points and how they are aligned without introducing new spatial bias) to allow verification that the alignment steps do not undermine the comparison to full-supervision upper bounds.

    Authors: We thank the referee for this suggestion to improve verifiability. While the original methods section outlines the high-level integration of LPO and UPO, we acknowledge that explicit loss equations and step-by-step construction of preference pairs from sparse points would strengthen the presentation. In the revised manuscript we have added the full mathematical formulations for both the LPO and UPO losses together with the algorithmic procedure for generating spatially localized preference pairs. These additions confirm that the alignment process preserves spatial representativeness and does not introduce bias that would invalidate comparisons against full supervision. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical method with independent experimental validation

full rationale

The paper introduces Prefer-DAS as a promptable multitask model combining self-training, prompt-guided contrastive learning, LPO for local preference alignment, and UPO for unsupervised cases. These are presented as novel plug-and-play additions to handle sparse points and human feedback in target domains for DAS. Performance claims rest on comprehensive experiments across four tasks, with direct comparisons to SAM-like, UDA, weakly-supervised, and supervised baselines. No equations, derivations, or self-citation chains are invoked that reduce any result or prediction to fitted inputs or prior outputs by construction. The approach is self-contained, drawing on established techniques while adding verifiable new components evaluated externally via benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

The central claims rest on the effectiveness of newly introduced LPO and UPO modules plus the assumption that sparse prompts and local preferences transfer usefully across EM domains.

axioms (1)
  • domain assumption Local human preferences provide spatially varying feedback that can be directly optimized for segmentation alignment
    Invoked when introducing LPO as a plug-and-play solution for alignment with human feedback
invented entities (2)
  • Local direct Preference Optimization (LPO) no independent evidence
    purpose: Align model outputs with spatially varying human feedback in the target domain
    New optimization technique presented as the core of preference alignment
  • Unsupervised Preference Optimization (UPO) no independent evidence
    purpose: Generate and leverage self-learned preferences when human feedback is missing
    Introduced to handle cases with no human input while maintaining the preference framework

pith-pipeline@v0.9.0 · 5818 in / 1497 out tokens · 47610 ms · 2026-05-21T13:11:57.753396+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

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

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

39 extracted references · 39 canonical work pages

  1. [1]

    Call to action to properly utilize electron microscopy to measure organelles to monitor disease,

    K. Neikirk, E.-G. Lopez, A. G. Marshall, A. Alghanem, E. Krystofiak, B. Kula, N. Smith, J. Shao, P. Katti, and A. Hinton Jr, “Call to action to properly utilize electron microscopy to measure organelles to monitor disease,”European Journal of Cell Biology, vol. 102, no. 4, p. 151365, 2023

    work page 2023

  2. [2]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention, 2015, pp. 234–241

    work page 2015

  3. [3]

    An image is worth 16x16 words: Trans- formers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Trans- formers for image recognition at scale,” inInternational Conference on Learning Representations, 2021

    work page 2021

  4. [4]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Loet al., “Segment anything,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 4015–4026

    work page 2023

  5. [5]

    A review of the segment anything model (sam) for medi- cal image analysis: Accomplishments and perspectives,

    M. Ali, T. Wu, H. Hu, Q. Luo, D. Xu, W. Zheng, N. Jin, C. Yang, and J. Yao, “A review of the segment anything model (sam) for medi- cal image analysis: Accomplishments and perspectives,”Computerized Medical Imaging and Graphics, vol. 119, p. 102473, 2025

    work page 2025

  6. [6]

    Improving the generalization of segmentation foundation model under distribution shift via weakly supervised adaptation,

    H. Zhang, Y . Su, X. Xu, and K. Jia, “Improving the generalization of segmentation foundation model under distribution shift via weakly supervised adaptation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 23 385–23 395

    work page 2024

  7. [7]

    Sam-med2d

    J. Cheng, J. Ye, Z. Deng, J. Chen, T. Li, H. Wang, Y . Su, Z. Huang, J. Chen, L. Jianget al., “Sam-med2d,”arXiv preprint arXiv:2308.16184, 2023. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13 TABLE V INFLUENCE OF THE QUANTITY OF INFERENCE POINT PROMPTS ON INTERACTIVE SEGMENTATION. WHILESAM-LIKE METHODS REQUIRES FULL POINT PROMPTS FOR INTERAC...

    work page arXiv 2023

  8. [8]

    Medical sam adapter: Adapting segment anything model for medical image segmentation,

    J. Wu, W. Ji, Y . Liu, H. Fu, M. Xu, Y . Xu, and Y . Jin, “Medical sam adapter: Adapting segment anything model for medical image segmentation,”Medical Image Analysis, vol. 102, p. 103547, 2025

    work page 2025

  9. [9]

    Deep reinforcement learning from human preferences,

    P. F. Christiano, J. Leike, T. Brown, M. Martic, S. Legg, and D. Amodei, “Deep reinforcement learning from human preferences,”Advances in Neural Information Processing Systems, vol. 30, 2017

    work page 2017

  10. [10]

    Direct preference optimization: Your language model is secretly a reward model,

    R. Rafailov, A. Sharma, E. Mitchell, C. D. Manning, S. Ermon, and C. Finn, “Direct preference optimization: Your language model is secretly a reward model,”Advances in Neural Information Processing Systems, vol. 36, pp. 53 728–53 741, 2023

    work page 2023

  11. [11]

    Diffusion model alignment using direct preference optimization,

    B. Wallace, M. Dang, R. Rafailov, L. Zhou, A. Lou, S. Purushwalkam, S. Ermon, C. Xiong, S. Joty, and N. Naik, “Diffusion model alignment using direct preference optimization,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 8228–8238

    work page 2024

  12. [12]

    Aligning diffusion models by optimizing human utility,

    S. Li, K. Kallidromitis, A. Gokul, Y . Kato, and K. Kozuka, “Aligning diffusion models by optimizing human utility,”Advances in Neural Information Processing Systems, vol. 37, pp. 24 897–24 925, 2024

    work page 2024

  13. [13]

    Enhancing sam with efficient prompting and preference optimiza- tion for semi-supervised medical image segmentation,

    A. Konwer, Z. Yang, E. Bas, C. Xiao, P. Prasanna, P. Bhatia, and T. Kass- Hout, “Enhancing sam with efficient prompting and preference optimiza- tion for semi-supervised medical image segmentation,” inProceedings of the Computer Vision and Pattern Recognition, 2025, pp. 20 990–21 000

    work page 2025

  14. [14]

    Causal confusion and reward misidentification in preference-based reward learning,

    J. Tien, J. Z.-Y . He, Z. Erickson, A. D. Dragan, and D. S. Brown, “Causal confusion and reward misidentification in preference-based reward learning,” inThe Eleventh International Conference on Learning Representations, 2023

    work page 2023

  15. [15]

    Prompt-das: Annotation-efficient prompt learning for domain adaptive semantic segmentation of electron microscopy images,

    J. Chen, S. Xiong, and J. Peng, “Prompt-das: Annotation-efficient prompt learning for domain adaptive semantic segmentation of electron microscopy images,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2025, pp. 512–521

    work page 2025

  16. [16]

    Cycada: Cycle-consistent adversarial domain adapta- tion,

    J. Hoffman, E. Tzeng, T. Park, J.-Y . Zhu, P. Isola, K. Saenko, A. Efros, and T. Darrell, “Cycada: Cycle-consistent adversarial domain adapta- tion,” inInternational Conference on Machine Learning. Pmlr, 2018, pp. 1989–1998

    work page 2018

  17. [17]

    Domain-adversarial training of neural networks,

    Y . Ganin, E. Ustinova, H. Ajakan, P. Germain, H. Larochelle, F. Lavi- olette, M. March, and V . Lempitsky, “Domain-adversarial training of neural networks,”Journal of Machine Learning Research, vol. 17, no. 59, pp. 1–35, 2016

    work page 2016

  18. [18]

    Learning to adapt structured output space for semantic segmentation,

    Y . Tsai, W. Hung, S. Schulter, K. Sohn, M. Yang, and M. Chandraker, “Learning to adapt structured output space for semantic segmentation,” inIEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 7472–7481

    work page 2018

  19. [19]

    Unsupervised mitochondria segmentation in em images via domain adaptive multi-task learning,

    J. Peng, J. Yi, and Z. Yuan, “Unsupervised mitochondria segmentation in em images via domain adaptive multi-task learning,”IEEE Journal of Selected Topics in Signal Processing, vol. 14, no. 6, pp. 1199–1209, 2020

    work page 2020

  20. [20]

    Uncertainty-aware label rectification for domain adaptive mitochondria segmentation,

    S. Wu, C. Chen, Z. Xiong, X. Chen, and X. Sun, “Uncertainty-aware label rectification for domain adaptive mitochondria segmentation,” in International Conference on Medical Image Computing and Computer Assisted Intervention, 2021, pp. 191–200

    work page 2021

  21. [21]

    Class-aware feature alignment for domain adaptative mitochondria segmentation,

    D. Yin, W. Huang, Z. Xiong, and X. Chen, “Class-aware feature alignment for domain adaptative mitochondria segmentation,” inInterna- tional Conference on Medical Image Computing and Computer-Assisted Intervention, 2023, pp. 238–248

    work page 2023

  22. [22]

    Weakly-supervised cross-domain segmentation of electron microscopy with sparse point annotation,

    D. Qiu, S. Xiong, J. Yi, and J. Peng, “Weakly-supervised cross-domain segmentation of electron microscopy with sparse point annotation,” IEEE Transactions on Big Data, vol. 11, no. 2, pp. 359–371, 2025

    work page 2025

  23. [23]

    Rank analysis of incomplete block designs: I. the method of paired comparisons,

    R. A. Bradley and M. E. Terry, “Rank analysis of incomplete block designs: I. the method of paired comparisons,”Biometrika, vol. 39, no. 3/4, pp. 324–345, 1952

    work page 1952

  24. [24]

    On softmax direct preference optimization for recommen- dation,

    Y . Chen, J. Tan, A. Zhang, Z. Yang, L. Sheng, E. Zhang, X. Wang, and T.-S. Chua, “On softmax direct preference optimization for recommen- dation,”Advances in Neural Information Processing Systems, vol. 37, pp. 27 463–27 489, 2024

    work page 2024

  25. [25]

    The analysis of permutations,

    R. L. Plackett, “The analysis of permutations,”Journal of the Royal Statistical Society Series C: Applied Statistics, vol. 24, no. 2, pp. 193– 202, 1975

    work page 1975

  26. [26]

    Popen: Preference-based optimization and ensemble for lvlm-based reasoning segmentation,

    L. Zhu, T. Chen, Q. Xu, X. Liu, D. Ji, H. Wu, D. W. Soh, and J. Liu, “Popen: Preference-based optimization and ensemble for lvlm-based reasoning segmentation,” inProceedings of the Computer Vision and Pattern Recognition, 2025, pp. 30 231–30 240

    work page 2025

  27. [27]

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

    A. Tarvainen and H. Valpola, “Mean teachers are better role mod- els: Weight-averaged consistency targets improve semi-supervised deep learning results,” inAdvances in Neural Information Processing Systems, vol. 30, 2017

    work page 2017

  28. [28]

    Dn-detr: Ac- celerate detr training by introducing query denoising,

    F. Li, H. Zhang, S. Liu, J. Guo, L. M. Ni, and L. Zhang, “Dn-detr: Ac- celerate detr training by introducing query denoising,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 13 619–13 627

    work page 2022

  29. [29]

    Distance regularized level set evolution and its application to image segmentation,

    C. Li, C. Xu, C. Gui, and M. D. Fox, “Distance regularized level set evolution and its application to image segmentation,”IEEE Transactions on Image Processing, vol. 19, no. 12, pp. 3243–3254, 2010

    work page 2010

  30. [30]

    Automatic abdominal multi-organ segmentation using deep convolutional neural network and time-implicit level sets,

    P. Hu, F. Wu, J. Peng, Y . Bao, F. Chen, and D. Kong, “Automatic abdominal multi-organ segmentation using deep convolutional neural network and time-implicit level sets,”International Journal of Computer Assisted Radiology and Surgery, vol. 12, no. 3, pp. 399–411, 2017

    work page 2017

  31. [31]

    Fast mitochondria segmentation for connectomics,

    V . Casser, K. Kang, H. Pfister, and D. Haehn, “Fast mitochondria segmentation for connectomics,” inProceedings of Machine Learning Research, 2020, pp. 111–120

    work page 2020

  32. [32]

    Supervoxel- based segmentation of mitochondria in em image stacks with learned shape features,

    A. Lucchi, K. Smith, R. Achanta, G. Knott, and P. Fua, “Supervoxel- based segmentation of mitochondria in em image stacks with learned shape features,”IEEE Transactions on Medical Imaging, vol. 31, no. 2, pp. 474–486, 2011

    work page 2011

  33. [33]

    Mitoem dataset: Large-scale 3d mitochondria instance segmentation from em images,

    D. Wei, Z. Lin, D. Franco-Barranco, N. Wendt, X. Liu, W. Yin, X. Huang, A. Gupta, W.-D. Jang, X. Wanget al., “Mitoem dataset: Large-scale 3d mitochondria instance segmentation from em images,” in International Conference on Medical Image Computing and Computer- Assisted Intervention, 2020, pp. 66–76

    work page 2020

  34. [34]

    A connectomic study of a petascale fragment of human cerebral cortex,

    A. Shapson-Coe, M. Januszewski, D. R. Berger, A. Pope, Y . Wu, T. Blakely, R. L. Schalek, P. H. Li, S. Wang, J. Maitin-Shepardet al., “A connectomic study of a petascale fragment of human cerebral cortex,” BioRxiv, pp. 2021–05, 2021

    work page 2021

  35. [35]

    Efficient cell-wide mapping of mitochondria in electron microscopic volumes using webknossos,

    Y . Jiang, H. Wang, K. M. Boergens, N. Rzepka, F. Wang, and Y . Hua, “Efficient cell-wide mapping of mitochondria in electron microscopic volumes using webknossos,”Cell Reports Methods, vol. 5, no. 2, 2025

    work page 2025

  36. [36]

    Domain adaptive mitochondria segmentation via enforcing inter-section consistency,

    W. Huang, X. Liu, Z. Cheng, Y . Zhang, and Z. Xiong, “Domain adaptive mitochondria segmentation via enforcing inter-section consistency,” in International Conference on Medical Image Computing and Computer- Assisted Intervention, 2022, pp. 89–98

    work page 2022

  37. [37]

    A dataset and a technique for generalized nuclear segmen- tation for computational pathology,

    N. Kumar, R. Verma, S. Sharma, S. Bhargava, A. Vahadane, and A. Sethi, “A dataset and a technique for generalized nuclear segmen- tation for computational pathology,”IEEE Transactions on Medical Imaging, vol. 36, no. 7, pp. 1550–1560, 2017

    work page 2017

  38. [38]

    Panoptic seg- mentation,

    A. Kirillov, K. He, R. Girshick, C. Rother, and P. Dollar, “Panoptic seg- mentation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 9404–9413. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 14

    work page 2019

  39. [39]

    Emerging properties in self-supervised vision transformers,

    M. Caron, H. Touvron, I. Misra, H. J ´egou, J. Mairal, P. Bojanowski, and A. Joulin, “Emerging properties in self-supervised vision transformers,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 9650–9660

    work page 2021