Pith. sign in

REVIEW 3 major objections 4 minor 49 references

Local2Global query Alignment for Video Instance Segmentation

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Early alignment of propagated global queries with current-frame local queries, via the lightweight L2G-aligner, is claimed to be the key to temporally consistent online video instance segmentation.

desk verdict A genuinely simple, well-ablated fix for temporal drift in online VIS that deserves peer review, but the local-query selection step is under-specified enough that the main result isn't yet reproducible. read the letter →

arxiv 2507.20120 v1 pith:YKE47OB5 submitted 2025-07-27 cs.CV

Rajat Koner , Zhipeng Wang , Srinivas Parthasarathy , Chinghang Chen This is my paper · ORCID
classification cs.CV
keywords videoinstancesegmentationquerypropagationtemporalconsistencyL2G-alignerDETRobjectqueriestransformerdecodertrajectorypositionembeddingonline
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

Local2Global's central claim is that in online video instance segmentation, temporally consistent predictions come from aligning the propagated global queries, which carry past object representations, with local queries drawn from the current frame before the segmentation decoder runs, rather than from adding memory banks or separate tracking modules. The paper introduces a lightweight transformer decoder, the L2G-aligner, that performs this alignment, and a trajectory position embedding computed from the previous frame's mask. It reports that adding the aligner raises AP on OVIS from 31.6 to 36.2 and on YouTube-VIS 2021 from 43.9 to 48.9, and that including trajectory propagation reaches 37.0 AP on OVIS and 54.3 AP on YouTube-VIS 2019 with a ResNet-50 backbone. A sympathetic reader would care because the method works with simple online training and no external memory, suggesting that the bottleneck in propagation-based VIS is the quality of the alignment step, not the complexity of the temporal model.

What carries the argument

The L2G-aligner is a lightweight transformer decoder (three layers, hidden dimension 1024, about 3.2 million parameters) placed between the segmentation encoder and decoder. It takes the global queries, which are the final decoder output queries from the previous frame and carry instance identity, as queries, and takes local queries, the current frame's top-K encoder tokens ranked by classification confidence with static learnable position embeddings, as key-value pairs; cross-attention then reorients the past representation toward the current frame. The second mechanism is the trajectory position embedding: the previous frame's predicted mask is converted to a box, encoded with a sinusoidal positional encoding, and projected by an MLP, so the global query's position is updated dynamically along the instance's path. Together the two mechanisms carry the argument: the aligner supplies current visual context, the dynamic position supplies current spatial context, and the decoder then applies its image-domain localization bias.

What would settle it

Measure the recall of the top-K selected local queries against ground-truth instance masks on OVIS frames with heavy occlusion; if instances that later reappear are absent from the selected token set in the frames where they are occluded, then the aligner cannot be the source of their recovery, and the paper's account of how current-frame cues enter the propagated query would fail that test.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that query propagation in DETR-style video segmentation degrades because past queries are handed to the segmentation decoder with a spatial bias trained for current-frame images, and the fix is to align those propagating global queries with a compact set of current-frame local queries (the top-K encoder tokens ranked by classification confidence) through a lightweight cross-attention decoder placed between encoder and decoder. Local queries carry the current frame's visual cues without committing to a spatial anchor, global queries maintain instance identity across frames, and the alignment reorients the past representation before the decoder uses it. Explicit trajectory information is then added by computing dynamic position embeddings from the previous frame's mask box, so the query both looks and moves like the instance. The ablations attribute 4.6 to 5.0 AP of the gain to the aligner and a further 0.8 AP on OVIS to trajectory embeddings, leading to 54.3 and 49.4 AP on YouTube-VIS 2019 and 2021, 42.3 AP on YouTube-VIS 2022, 37.0 AP on OVIS, and 46.3 VPQ on VIPSeg.

Load-bearing premise

The method assumes that the few current-frame feature entries with the highest classification scores are a faithful and sufficient summary of every object in the frame, including occluded and never-before-seen objects; if that selection misses an instance, the aligner can only inject noise into that instance's global query.

Editorial extensions

If this is right

  • Propagating raw queries through the decoder without alignment is the identified failure mode; the paper's baseline loses 4.6 to 5.0 AP compared with the same model using the L2G-aligner.
  • Online video instance segmentation does not need separate memory banks, decoupled temporal transformers, or offline-then-online training to match or beat offline methods on the tested benchmarks.
  • The segmentation decoder's pretrained frame-level bias is a reusable asset for video, provided past queries are re-anchored to current-frame cues before decoding.
  • Explicitly propagating where an instance was, through dynamic position embeddings from the previous mask, is worth about 0.8 AP on the heavily occluded OVIS dataset.
  • Static rather than dynamic position embeddings for local queries is the right design choice, because unrefined local tokens contain duplicate and background predictions.

Reading between the lines

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

  • A testable extension is replacing the confidence-based top-K selection with a learned or mask-guided proposal step; the paper's own limitation section reports failures when objects disappear and reappear, which is exactly the regime where confidence-selected tokens are most likely to miss the object.
  • If the alignment hypothesis generalizes, the same early-alignment step should transfer to other DETR-family propagation pipelines, such as video object detection or video object segmentation, with a similar reduction in drift.
  • The static-position choice for local queries suggests a broader design principle for query-based video models: do not attach dynamic, instance-specific position embeddings to unrefined candidate tokens; apply them only after alignment has anchored the token.
  • Stress-testing the claim on a dataset with heavy occlusion and object reappearance, such as the MOSE dataset the paper itself names, would locate the boundary of the alignment 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 / 4 minor

Summary. The paper proposes Local2Global, an online video instance segmentation framework built on Mask2Former with query propagation. It introduces two query sets: local queries, selected from current-frame encoder features by token-level classification confidence, and global queries, which are the propagated decoder queries from past frames. A lightweight transformer decoder, the L2G-aligner, performs cross-attention with the local queries as key/value and global queries as queries, producing an aligned query representation before the segmentation decoder. The method also propagates trajectory position embeddings computed from the previous frame's predicted masks via Mask2Box, sinusoidal encoding, and an MLP. Training uses standard detection/segmentation losses plus a reduced-supervision scheme that randomly skips intermediate frames when computing losses. Experiments report state-of-the-art results on YouTube-VIS 2019/2021/2022, OVIS, and VIPSeg; the main ablation (Table 3) shows the L2G-aligner adds 4.6 AP on OVIS and 5.0 AP on YTVIS-2021 over the propagated-query baseline, and trajectory embeddings add another 0.8 AP on OVIS.

Significance. If the claims hold, the contribution is significant for online video instance segmentation: a simple, lightweight module integrated directly into the segmentation decoder path improves temporal consistency without external memory banks, separate tracking modules, or multi-stage offline-to-online training. The ablation in Table 3 directly isolates the proposed L2G-aligner and shows a large and consistent gain across two datasets, which is the paper's central evidence. The cross-dataset evaluation (YTVIS-19/21/22, OVIS, VIPSeg) is broad, and the reported parameter overhead (3.2M parameters) is modest. However, the current manuscript has a load-bearing reproducibility gap: the local-query selection mechanism is not fully specified, and the reported YTVIS-2019 number conflicts between the abstract/table and the main text. The absence of code and error bars further tempers confidence, although these are not themselves disqualifying given the size of the reported effects.

major comments (3)
  1. [Section 3.2 (L2G-aligner paragraph, Eq. (3))] The local-query selection mechanism is undefined. The text states that qlocal is obtained from the encoder E "through the topK∈ N×c token prediction, where the topK is determined based on the classification confidence computed by the class-prediction head for each token." In the Mask2Former backbone described in Section 3.1, the class-prediction head operates on the decoder output queries, not on the encoder's spatial tokens. The paper does not define a new token-level classification head, nor its loss, training schedule, or the value of K, nor how the head is applied to the multi-scale encoder features. Because qlocal serves as the key/value input to the L2G-aligner in Eq. (3), the forward computation of the central proposed module cannot be reproduced from the paper as written. Please provide the full specification of this token-classification head, its supervision, K, and the token selection procedure, or replace the mechanism with one that is fully derivable from the described architecture.
  2. [Section 4.3 (YTVIS-2019/2021) vs. Abstract and Table 1] There is a numerical inconsistency in the headline result. The abstract and Table 1 report 54.3 AP on YouTube-VIS-2019, but Section 4.3 states "Local2Global achieves 53.4 AP" and claims it outperforms the closest online method by 3.1 AP. The 3.1 AP margin is consistent only with 54.3 AP against DVIS (51.2 AP), not with 53.4 AP. This discrepancy must be corrected and the correct number used consistently throughout.
  3. [Table 3 and Section 4.4 (Effect of the proposed L2G-aligner)] The ablation baseline's use of position embeddings is ambiguous, which matters for interpreting the central claim. Section 3.1 introduces anchor-box dynamic positional embeddings as part of the backbone and states they act as trajectory embeddings, yet Table 3 treats "trajectory" as a separate addition after the L2G-aligner. It is therefore unclear whether each row of Table 3 uses static or dynamic position embeddings for the global queries, and whether the baseline already includes the dynamic position embeddings from Eq. (2). Please state explicitly, row by row, which positional embeddings are used, so that the +4.6/+5.0 AP gain attributed to the L2G-aligner is not confounded with the gain from switching to trajectory position embeddings.
minor comments (4)
  1. [Section 3.1] There are several typos: "both both positional and visual features" and "instance of an instance" should be corrected, and the phrase "it comprising" is ungrammatical.
  2. [Section 3.2] The word "segementation" is misspelled; also the expression "topK∈ N×c token prediction" is unclear notation and should be rewritten, e.g., as "the top-K tokens are selected and projected to a set of N query vectors of dimension c."
  3. [Section 4.2] The paper states all experiments use a single training run without reporting variance. Please indicate the number of random seeds and, if only one run was performed, state this explicitly in the experimental setup.
  4. [Abstract and Section 1] The phrase "without bells and rings" is informal; consider replacing with "without bells and whistles" or a standard phrase such as "with a simple architecture."

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claim is supported by ablations on external benchmarks, and the only overlapping-author citation is not load-bearing.

full rationale

The paper's central claim is that early alignment of past global queries with current-frame local queries improves temporal consistency in online video instance segmentation. The support for this claim is empirical: Table 3 compares a baseline that only propagates global queries through the segmentation decoder against the same model plus the L2G-aligner, measuring AP on OVIS and YouTube-VIS-2021, both external benchmarks that are not used to fit any constant in the method. Those ablations constitute standard model selection rather than circular reasoning, since the design choice is validated against held-out evaluation data and against the stated baseline. The local queries are selected as topK encoder tokens by classification confidence; although the paper does not specify how that class-prediction head is defined or trained, which is a reproducibility gap, there is no equation that identifies qlocal with the final predictions or otherwise makes the alignment result an input by construction. The only citation with overlapping authorship is InstanceFormer [24], cited alongside GenVIS [18] as prior query-propagation work; the contribution of this paper, the L2G-aligner and trajectory position embeddings, is not derived from or justified by that citation, so the self-citation is not load-bearing. No uniqueness theorem is imported from the authors' prior work, no ansatz is smuggled in through a self-citation, and no known result is merely renamed. Accordingly, no circular step can be exhibited with the specificity required by the review rules.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central claim rests on standard deep-learning assumptions (transfer from COCO, query identity, reliable encoder token selection) and on a set of hand-chosen hyperparameters (aligner dimension, layer count, topK count, frame-skip schedule). No new physical or mathematical entities are postulated.

free parameters (4)
  • L2G-aligner hidden dimension = 1024
    Selected by validation ablation Table 5, tested at 512, 1024, and 2048.
  • Number of L2G-aligner decoder layers = 3
    Stated in Section 4.2; not ablated.
  • topK local query count = not reported
    Local queries are selected as the topK encoder tokens by classification confidence in Section 3.2; the exact K is not specified.
  • Reduced supervision frame-skip schedule = random skip of 1 to 3 of 4 frames
    Introduced in Section 3.2 as a training heuristic and ablated in Table 6 with a gain of 0.6 AP on OVIS.
assumptions (6)
  • standard math Transformer cross-attention is a working operation.
    Used in the L2G-aligner and segmentation decoder without proof.
  • standard math Hungarian matching provides valid one-to-one assignment.
    Used in Section 3.2 'Overall Loss' for matching predictions to ground truth.
  • domain assumption COCO pre-training transfers to video datasets.
    L2G-aligner is pre-trained on COCO and video training includes COCO-generated videos in Section 4.2.
  • domain assumption Propagated queries maintain instance identity across frames.
    The matching is computed once per ground truth instance as soon as it appears in Section 3.2.
  • domain assumption TopK encoder tokens selected by class confidence faithfully represent current-frame instances.
    qlocal is derived this way in Section 3.2, and this is the basis for local query construction.
  • ad hoc to paper Reduced supervision (randomly skipping frames for loss computation) improves training.
    Proposed in Section 3.2 as a training technique, not derived from first principles.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Local2Global query Alignment for Video Instance Segmentation." pith.science (2026). https://pith.science/paper/YKE47OB5

@misc{pith2026250720120,
  author       = {Pith},
  title        = {Pith review of: Local2Global query Alignment for Video Instance Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YKE47OB5}},
  note         = {Machine review of arXiv:2507.20120}
}
read the original abstract

Online video segmentation methods excel at handling long sequences and capturing gradual changes, making them ideal for real-world applications. However, achieving temporally consistent predictions remains a challenge, especially with gradual accumulation of noise or drift in on-line propagation, abrupt occlusions and scene transitions. This paper introduces Local2Global, an online framework, for video instance segmentation, exhibiting state-of-the-art performance with simple baseline and training purely in online fashion. Leveraging the DETR-based query propagation framework, we introduce two novel sets of queries:(1) local queries that capture initial object-specific spatial features from each frame and (2) global queries containing past spatio-temporal representations. We propose the L2G-aligner, a novel lightweight transformer decoder, to facilitate an early alignment between local and global queries. This alignment allows our model to effectively utilize current frame information while maintaining temporal consistency, producing a smooth transition between frames. Furthermore, L2G-aligner is integrated within the segmentation model, without relying on additional complex heuristics, or memory mechanisms. Extensive experiments across various challenging VIS and VPS datasets showcase the superiority of our method with simple online training, surpassing current benchmarks without bells and rings. For instance, we achieve 54.3 and 49.4 AP on Youtube-VIS-19/-21 datasets and 37.0 AP on OVIS dataset respectively withthe ResNet-50 backbone.

Figures

Figures reproduced from arXiv: 2507.20120 by the authors.

Figure 1
Figure 1. Overall architecture of the proposed Local2Global framework(left). Instance queries with color represent their corresponding [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Qualitative examples of our model. In (a) our Local2Global framework successfully segments and tracks a large number of fish [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Qualitative evaluation. Columns from left to right are [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 42 canonical work pages

  1. [1]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion

    Athar, A., Hermans, A., Luiten, J., Ramanan, D., Leibe, B.: Tarvis: A unified approach for target-based video segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion. pp. 18738–18748 (2023) 7

    work page 2023

  2. [2]

    In: ECCV (2020) 3

    Athar, A., Mahadevan, S., Osep, A., Leal-Taix ´e, L., Leibe, B.: Stem-seg: Spatio-temporal embeddings for instance segmentation in videos. In: ECCV (2020) 3

    work page 2020

  3. [3]

    In: CVPR (2020) 1

    Bertasius, G., Torresani, L.: Classifying, segmenting, and tracking object instances in video with mask prop- agation. In: CVPR (2020) 1

    work page 2020

  4. [4]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Cai, J., Xu, M., Li, W., Xiong, Y ., Xia, W., Tu, Z., Soatto, S.: Memot: Multi-object tracking with mem- ory. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 8090– 8100 (2022) 3

    work page 2022

  5. [5]

    In: ECCV (2020) 2

    Cao, J., Anwer, R.M., Cholakkal, H., Khan, F.S., Pang, Y ., Shao, L.: Sipmask: Spatial information preservation for fast image and video instance seg- mentation. In: ECCV (2020) 2

    work page 2020

  6. [6]

    In: ECCV (2020) 1, 2, 8

    Carion, N., Massa, F., Synnaeve, G., Usunier, N., Kir- illov, A., Zagoruyko, S.: End-to-end object detection with transformers. In: ECCV (2020) 1, 2, 8

    work page 2020

  7. [7]

    In: Proceedings of the IEEE/CVF confer- ence on computer vision and pattern recognition

    Chen, Y ., Cao, Y ., Hu, H., Wang, L.: Memory en- hanced global-local aggregation for video object de- tection. In: Proceedings of the IEEE/CVF confer- ence on computer vision and pattern recognition. pp. 10337–10346 (2020) 2

    work page 2020

  8. [9]

    arXiv:2112.10764 (2021) 2, 4, 6

    Cheng, B., Choudhuri, A., Misra, I., Kirillov, A., Gird- har, R., Schwing, A.G.: Mask2former for video in- stance segmentation. arXiv:2112.10764 (2021) 2, 4, 6

    arXiv 2021

Show all 49 references
  1. [10]

    In: NeurIPS (2021) 1, 2, 3

    Cheng, B., Schwing, A.G., Kirillov, A.: Per-pixel classification is not all you need for semantic segmen- tation. In: NeurIPS (2021) 1, 2, 3

    2021

  2. [11]

    In: Proceedings of the IEEE/CVF in- ternational conference on computer vision

    Ding, H., Liu, C., He, S., Jiang, X., Torr, P.H., Bai, S.: Mose: A new dataset for video object segmentation in complex scenes. In: Proceedings of the IEEE/CVF in- ternational conference on computer vision. pp. 20224– 20234 (2023) 8

    2023

  3. [12]

    In: Proceedings of the IEEE/CVF conference on com- puter vision and pattern recognition

    Ge, W., Lu, X., Shen, J.: Video object segmenta- tion using global and instance embedding learning. In: Proceedings of the IEEE/CVF conference on com- puter vision and pattern recognition. pp. 16836–16845 (2021) 2

    2021

  4. [13]

    In: CVPR (2022) 1

    Han, S.H., Hwang, S., Oh, S.W., Park, Y ., Kim, H., Kim, M.J., Kim, S.J.: Visolo: Grid-based space-time aggregation for efficient online video instance seg- mentation. In: CVPR (2022) 1

    2022

  5. [14]

    arXiv preprint arXiv:2305.17096 (2023) 3

    Hannan, T., Koner, R., Bernhard, M., Shit, S., Menze, B., Tresp, V ., Schubert, M., Seidl, T.: Gratt-vis: Gated residual attention for auto rectifying video in- stance segmentation. arXiv preprint arXiv:2305.17096 (2023) 3

    2023 arXiv

  6. [15]

    arXiv preprint arXiv:2202.07025 (2022) 2

    Hannan, T., Koner, R., Kobold, J., Schubert, M.: Box supervised video segmentation proposal network. arXiv preprint arXiv:2202.07025 (2022) 2

    2022 arXiv

  7. [16]

    In: ICCV (2017) 1, 3

    He, K., Gkioxari, G., Doll ´ar, P., Girshick, R.: Mask r-cnn. In: ICCV (2017) 1, 3

    2017

  8. [17]

    In: CVPR (2016) 4

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: CVPR (2016) 4

    2016

  9. [18]

    In: CVPR (2023) 1, 2, 3, 4, 5, 6, 7, 11

    Heo, M., Hwang, S., Hyun, J., Kim, H., Oh, S.W., Lee, J.Y ., Kim, S.J.: A generalized framework for video instance segmentation. In: CVPR (2023) 1, 2, 3, 4, 5, 6, 7, 11

    2023

  10. [19]

    In: NeurIPS (2022) 1, 2, 3, 4, 5, 6

    Heo, M., Hwang, S., Oh, S.W., Lee, J.Y ., Kim, S.J.: Vita: Video instance segmentation via object token as- sociation. In: NeurIPS (2022) 1, 2, 3, 4, 5, 6

    2022

  11. [20]

    In: NeurIPS (2022) 2, 5

    Huang, D.A., Yu, Z., Anandkumar, A.: Minvis: A minimal video instance segmentation framework without video-based training. In: NeurIPS (2022) 2, 5

    2022

  12. [21]

    In: NeurIPS (2021) 1, 2

    Hwang, S., Heo, M., Oh, S.W., Kim, S.J.: Video in- stance segmentation using inter-frame communication transformers. In: NeurIPS (2021) 1, 2

    2021

  13. [22]

    In: European Conference on Computer Vision

    Jiang, Z., Gu, Z., Peng, J., Zhou, H., Liu, L., Wang, Y ., Tai, Y ., Wang, C., Zhang, L.: Stc: spatio-temporal contrastive learning for video instance segmentation. In: European Conference on Computer Vision. pp. 539–556. Springer (2022) 1

    2022

  14. [23]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Kim, D., Woo, S., Lee, J.Y ., Kweon, I.S.: Video panoptic segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 9859–9868 (2020) 1, 7

    2020

  15. [24]

    In: Pro- ceedings of the AAAI Conference on Artificial Intel- ligence

    Koner, R., Hannan, T., Shit, S., Sharifzadeh, S., Schu- bert, M., Seidl, T., Tresp, V .: Instanceformer: An on- line video instance segmentation framework. In: Pro- ceedings of the AAAI Conference on Artificial Intel- ligence. vol. 37, pp. 1188–1195 (2023) 1, 2, 4, 5

    2023

  16. [25]

    In: Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition

    Li, F., Zhang, H., Xu, H., Liu, S., Zhang, L., Ni, L.M., Shum, H.Y .: Mask dino: Towards a unified transformer-based framework for object detection and segmentation. In: Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition. pp. 3041–3050 (2023) 2, 4, 6

    2023

  17. [26]

    In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision

    Li, J., Yu, B., Rao, Y ., Zhou, J., Lu, J.: Tcovis: Tem- porally consistent online video instance segmentation. In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision. pp. 1097–1107 (2023) 2, 3, 4, 6, 7

    2023

  18. [27]

    IEEE transactions on pattern analysis and machine intelligence (2024) 1

    Li, X., Ding, H., Yuan, H., Zhang, W., Pang, J., Cheng, G., Chen, K., Liu, Z., Loy, C.C.: Transformer-based visual segmentation: A survey. IEEE transactions on pattern analysis and machine intelligence (2024) 1

    2024

  19. [28]

    Li, X., Yuan, H., Li, W., Ding, H., Wu, S., Zhang, W., Li, Y ., Chen, K., Loy, C.C.: Omg-seg: Is one model good enough for all segmentation? In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 27948–27959 (2024) 8

    2024

  20. [29]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vi- sion

    Li, X., Yuan, H., Zhang, W., Cheng, G., Pang, J., Loy, C.C.: Tube-link: A flexible cross tube framework for universal video segmentation. In: Proceedings of the IEEE/CVF International Conference on Computer Vi- sion. pp. 13923–13933 (2023) 7

    2023

  21. [30]

    In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li, X., Zhang, W., Pang, J., Chen, K., Cheng, G., Tong, Y ., Loy, C.C.: Video k-net: A simple, strong, and unified baseline for video segmentation. In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 18847–18857 (2022) 7

    2022

  22. [31]

    In: ECCV (2014) 4, 8

    Lin, T.Y ., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Doll ´ar, P., Zitnick, C.L.: Microsoft coco: Common objects in context. In: ECCV (2014) 4, 8

    2014

  23. [32]

    arXiv preprint arXiv:2201.12329 (2022) 2, 4, 6, 8

    Liu, S., Li, F., Zhang, H., Yang, X., Qi, X., Su, H., Zhu, J., Zhang, L.: Dab-detr: Dynamic an- chor boxes are better queries for detr. arXiv preprint arXiv:2201.12329 (2022) 2, 4, 6, 8

    2022 arXiv

  24. [33]

    ICCV (2021) 8

    Liu, Z., Lin, Y ., Cao, Y ., Hu, H., Wei, Y ., Zhang, Z., Lin, S., Guo, B.: Swin transformer: Hierarchical vi- sion transformer using shifted windows. ICCV (2021) 8

    2021

  25. [34]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition

    Miao, J., Wang, X., Wu, Y ., Li, W., Zhang, X., Wei, Y ., Yang, Y .: Large-scale video panoptic segmenta- tion in the wild: A benchmark. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition. pp. 21033–21043 (2022) 6, 7

    2022

  26. [35]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (June 2020) 2

    Mun, J., Cho, M., Han, B.: Local-global video-text interactions for temporal grounding. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (June 2020) 2

    2020

  27. [36]

    IJCV (2022) 1, 6

    Qi, J., Gao, Y ., Hu, Y ., Wang, X., Liu, X., Bai, X., Belongie, S., Yuille, A., Torr, P.H., Bai, S.: Oc- cluded video instance segmentation: A benchmark. IJCV (2022) 1, 6

    2022

  28. [37]

    arXiv preprint arXiv:2408.00714 (2024), https:// arxiv.org/abs/2408.00714 8

    Ravi, N., Gabeur, V ., Hu, Y .T., Hu, R., Ryali, C., Ma, T., Khedr, H., R¨adle, R., Rolland, C., Gustafson, L., Mintun, E., Pan, J., Alwala, K.V ., Carion, N., Wu, C.Y ., Girshick, R., Doll ´ar, P., Feichtenhofer, C.: Sam 2: Segment anything in images and videos. arXiv preprin...

    2024 arXiv

  29. [38]

    Ren, T., Liu, S., Li, F., Zhang, H., Zeng, A., Yang, J., Liao, X., Jia, D., Li, H., Cao, H., Wang, J., Zeng, Z., Qi, X., Yuan, Y ., Yang, J., Zhang, L.: detrex: Bench- marking detection transformers (2023) 6

    2023

  30. [39]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision

    Shin, I., Kim, D., Yu, Q., Xie, J., Kim, H.S., Green, B., Kweon, I.S., Yoon, K.J., Chen, L.C.: Video- kmax: A simple unified approach for online and near- online video panoptic segmentation. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. ...

    2024

  31. [40]

    In: NeurIPS (2017) 1, 2, 4

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, Ł., Polosukhin, I.: Attention is all you need. In: NeurIPS (2017) 1, 2, 4

    2017

  32. [41]

    In: CVPR (2021) 2

    Wang, Y ., Xu, Z., Wang, X., Shen, C., Cheng, B., Shen, H., Xia, H.: End-to-end video instance segmen- tation with transformers. In: CVPR (2021) 2

    2021

  33. [42]

    In: ECCV (2022) 1, 5

    Wu, J., Jiang, Y ., Bai, S., Zhang, W., Bai, X.: Seq- former: Sequential transformer for video instance seg- mentation. In: ECCV (2022) 1, 5

    2022

  34. [43]

    In: ECCV (2022) 1, 2, 5, 6, 7

    Wu, J., Liu, Q., Jiang, Y ., Bai, S., Yuille, A., Bai, X.: In defense of online models for video instance seg- mentation. In: ECCV (2022) 1, 2, 5, 6, 7

    2022

  35. [44]

    In: ICCV (2019) 1, 2, 3, 6

    Yang, L., Fan, Y ., Xu, N.: Video instance segmenta- tion. In: ICCV (2019) 1, 2, 3, 6

    2019

  36. [45]

    In: ICCV (2021) 5

    Yang, S., Fang, Y ., Wang, X., Li, Y ., Fang, C., Shan, Y ., Feng, B., Liu, W.: Crossover learning for fast on- line video instance segmentation. In: ICCV (2021) 5

    2021

  37. [46]

    In: CVPR (2022) 5

    Yang, S., Wang, X., Li, Y ., Fang, Y ., Fang, J., Liu, W., Zhao, X., Shan, Y .: Temporally efficient vision trans- former for video instance segmentation. In: CVPR (2022) 5

    2022

  38. [47]

    In: European Conference on Com- puter Vision

    Zeng, F., Dong, B., Zhang, Y ., Wang, T., Zhang, X., Wei, Y .: Motr: End-to-end multiple-object tracking with transformer. In: European Conference on Com- puter Vision. pp. 659–675. Springer (2022) 3

    2022

  39. [48]

    arXiv:2203.03605 (2022) 4

    Zhang, H., Li, F., Liu, S., Zhang, L., Su, H., Zhu, J., Ni, L.M., Shum, H.Y .: Dino: Detr with improved de- noising anchor boxes for end-to-end object detection. arXiv:2203.03605 (2022) 4

    2022 arXiv

  40. [49]

    arXiv preprint arXiv:2306.03413 (2023) 1, 2, 4, 5, 6, 7

    Zhang, T., Tian, X., Wu, Y ., Ji, S., Wang, X., Zhang, Y ., Wan, P.: Dvis: Decoupled video instance segmen- tation framework. arXiv preprint arXiv:2306.03413 (2023) 1, 2, 4, 5, 6, 7

    2023 arXiv

  41. [50]

    Zhu, X., Su, W., Lu, L., Li, B., Wang, X., Dai, J.: Deformable detr: Deformable transformers for end- to-end object detection. In: ICLR (2020) 1, 4 Local2Global query Alignment: A few Qualitative Examples To provide a comprehensive understanding of our approach, we present a s...

    2020

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.