Pith. sign in

REVIEW 3 major objections 5 minor 69 references

SP$^2$T: Sparse Proxy Attention for Dual-stream Point Transformer

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read This paper claims that sparse proxy attention over grid-cell corners gives point transformers a global receptive field at near-linear cost, and that the dual-stream design beats PTv3 on indoor and outdoor 3D benchmarks.

desk verdict Solid empirical architecture paper; the sparse-attention normalization is stated ambiguously enough that the authors must clarify before the global-fusion claim is accepted. read the letter →

arxiv 2412.11540 v2 pith:FOHCPHLY submitted 2024-12-16 cs.CV cs.AI

classification cs.CVcs.AI
keywords pointcloud3Dsemanticsegmentationtransformersparseproxyattentiondual-streamnetworkrelativepositionbiasLiDARperceptionreceptivefield
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

Point transformers struggle to balance a large receptive field against computational cost and attention dilution. This paper proposes SP2T, a dual-stream transformer that runs a compressed proxy stream for global context alongside a point stream for local detail, exchanging features through sparse proxy attention at $O(k n d + m^2 d)$ cost with $k=8$ and $m$ much smaller than $n$. The paper claims this design achieves state-of-the-art results with acceptable latency: +3.8% mIoU over SPoTr on S3DIS, +22.9% over PointASNL on SemanticKITTI, and consistent gains over PTv3 on ScanNet, ScanNet200, S3DIS, nuScenes, SemanticKITTI, and Waymo. If correct, the work matters because it shows a path to very large receptive fields in point cloud transformers without quadratic attention cost or the attention dilution that comes from simply enlarging group windows.

What carries the argument

The load-bearing mechanism is sparse proxy attention (SPA). Each point is linked, by vertex-based association, to the eight corners of the containing axis-aligned grid cell, which is equivalent to L-infinity KNN with $k=8$; the attention then computes exponential similarities only within each proxy's association list, using map-reduce to obtain the sparse softmax and aggregate values. This reduces point–proxy interaction to $O(k n d)$ and leaves proxy–proxy global fusion at $O(m^2 d)$ with $m$ much smaller than $n$. A table-based relative bias (TRB) -- a trilinearly interpolated lookup table over relative positions, clamped and scaled -- injects spatial structure into the sparse similarities, and the paper's ablation attributes +1.20 mIoU on ScanNet to attention-based point–proxy interaction with TRB plus global fusion. The dual-stream layout, with PTv3 serialization attention handling local point fusion and TRB-driven self-attention among proxies handling global fusion, is what keeps local detail and global context balanced.

What would settle it

Take a held-out outdoor split (for example SemanticKITTI validation), replace the vertex-based association with exact L-infinity nearest-proxy association, and measure mIoU; if accuracy is essentially unchanged, the grid-corner assumption is not what carries the reported gains, while a large drop when grid spacing is doubled would confirm it is load-bearing.

Watch

Extended reading notes

Core claim

The paper's central claim is that local proxies anchored to a spatial grid can give point transformers a global receptive field if three weaknesses of earlier proxy methods are fixed. Sampling must be spatial-wise: a binary search chooses the grid spacing so the proxy count lands in a target range, and each point is associated with the eight corners of the axis-aligned grid cell containing it, which is equivalent to L-infinity KNN with $k=8$. Interaction must be sparse: the point–proxy attention sums exponential similarities only within each proxy's association set, using a map-reduce pass, so each point touches only $k$ proxies instead of all $m$ proxies. Finally, local and global streams must run in parallel rather than being merged into one stream. With these fixes, SP2T reports state-of-the-art mIoU on ScanNet (78.7), ScanNet200 (37.0), S3DIS 6-fold (79.7), nuScenes (81.2), and SemanticKITTI (71.7), and improves over PTv3 on Waymo detection by 1.6 mAPH with a single input frame.

Load-bearing premise

The load-bearing premise is that every point is well represented by the eight corners of the axis-aligned grid cell containing it, so in sparse or uneven point clouds where those corners are empty or are not the nearest proxies, the global fusion quality depends on an association that may miss the true neighbors.

Editorial extensions

If this is right

  • The point–proxy interaction cost drops from $O(nmd)$ to $O(knd+m^2d)$ with $k=8$, making proxy-based global receptive fields practical for large indoor and outdoor scenes; the paper reports about 20 FPS on nuScenes.
  • SP2T improves over PTv3 on every evaluated benchmark: +1.2 mIoU on ScanNet, +1.8 on ScanNet200, +2.0 on S3DIS 6-fold, +1.6 mAPH on Waymo single-frame detection, and +0.8/+0.9 mIoU on nuScenes and SemanticKITTI.
  • Proxy-based point transformers can now be competitive on indoor and outdoor tasks simultaneously; earlier proxy methods like SPoTr and Fast PT lagged on indoor datasets, while PointASNL lagged outdoors.
  • The sparse proxy interaction is plug-and-play: attaching it to a Minkowski-convolution backbone lifts ScanNet mIoU by +3.5 and ScanNet200 by +5.3, matching the gains seen with the PTv3 backbone.
  • Empty grid cells, whose proxies have no associated points, still contribute when included in global fusion with zero-initialized features, improving accuracy and indicating that the absence of points carries spatial information.

Reading between the lines

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

  • The vertex-based association assumes the eight corners of an axis-aligned grid cell are good proxies for a point; in sparse outdoor LiDAR with large empty cells, those corners may not be the true nearest proxies, so a direct comparison against exact L-infinity KNN association would isolate how much of the reported gain depends on this geometric shortcut.
  • Because TRB shows stage-level similarity across layers, the learned relative bias may act as a transferable distance kernel; a testable extension would train TRB on one dataset and freeze it on another to see whether the spatial prior transfers.
  • The dual-stream design is not tied to PTv3; with the Minkowski-convolution backbone it already improves both ScanNet and ScanNet200, so sparse proxy attention could be grafted onto other point, voxel, or hybrid backbones for detection and tracking.
  • Since empty proxies improve accuracy, occupancy of the proxy grid itself is informative; an auxiliary loss predicting whether a proxy cell contains points could make the global fusion even more explicit.
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 SP2T, a dual-stream point transformer that adds a local-proxy stream to a point stream (PTv3-style local fusion) for indoor and outdoor 3D semantic segmentation, instance segmentation, and object detection. The main novelty is Sparse Proxy Attention (SPA), which connects each point to eight grid-vertex proxies and computes sparse attention with a map-reduce scheme, plus a Table-Based Relative Bias (TRB) term. The paper reports state-of-the-art or competitive results on ScanNet, ScanNet200, S3DIS, SemanticKITTI, nuScenes, and Waymo, with ablations on ScanNet validation supporting the proxy count, association count, sampling scheme, empty-proxy treatment, and TRB variants.

Significance. If correct, the core claim is important: a local-proxy dual-stream design can provide a global receptive field at O(k n d + m^2 d) cost (k=8, m << n), improving over strong PTv3 baselines by 1.2-2.0 mIoU on indoor segmentation and 1.6-1.8 mAPH on Waymo detection. The paper includes several strengths: a parameter-matched PTv3 comparison (Table 5 rows 1-2), a latency breakdown (Fig. 7), a discussion of test-time augmentation inconsistencies with a same-TTA comparison (Table 8), and a transfer study replacing PTv3 with MinkUNet (Table 7). These features make the empirical claims more credible than a pure SOTA table. However, the central methodological description in Sec. 3.3 has an indexing/normalization inconsistency that must be resolved before the architecture's mechanism can be evaluated, and the reported results are single-run numbers with hyperparameters selected on the same validation sets that are reported.

major comments (3)
  1. [Sec. 3.3, Eqs. (1)-(3)] The equations defining SPA are internally inconsistent, and the described operation is not a per-query sparse softmax as the text claims. In Eq. (2), W_i^h is normalized by summing over j with aspx_j = aspx_i, i.e., over all associations sharing the same proxy (key-side normalization), not over associations sharing the same point (query-side normalization). In Eq. (3), the output o_i^h is indexed by proxy i and aggregates point values v_{aspt_j}, which describes a point-to-proxy aggregation; but the surrounding text says the example transfers features from proxy to point, and Eq. (1) puts the point feature in the query (q_{aspt_i}) and the proxy feature in the key (k_{aspx_i}). Because of these swaps, the paper does not currently specify a consistent attention direction. If the implementation follows Eqs. (1)-(3), the operation is not a valid per-query softmax attention and the gains attributed to attn-based PPI w/ TRB in Table 5 rows 7-10 would not support the stated global-fusion mechanism. Please correct the indexing, define the normalization group (per query point or per query proxy), and show the associated map-reduce pseudocode.
  2. [Sec. 4.2, Tables 5, 6, and 10] All structural and hyperparameter ablations - proxy count, association count, sampling method, empty-proxy handling, TRB scales, and temperature - are tuned and reported on ScanNet validation, and the main claims rest on single runs without error bars or repeated seeds. Several reported differences are small relative to typical run-to-run variance: e.g., Table 5 rows 5 and 7 differ by 0.13 mIoU, and Table 6 rows 1-4 span 0.67 mIoU across proxy counts. Because the headline SOTA claims and the ablations are single numbers, I cannot assess whether the improvements are statistically reliable. Please provide multiple seeds with mean and standard deviation for the main results and key ablations, or at minimum state the number of runs; if retraining is infeasible, report at least the held-out test numbers for the configurations compared in Table 5 and Table 10.
  3. [Sec. 3.3, Eq. (5)] The TRB formula appears inconsistent with the standard relative-bias mechanism described in the text. Eq. (5) writes S_i^h = exp(q·k/√d) + TRB^h(x), which adds the bias after the exponential, so the softmax denominator sums terms of the form exp(score) + bias. This is not the usual logit-bias form exp(q·k/√d + b) that Swin-style relative bias uses, and it changes the normalization statistics. If the implementation uses logit addition, please rewrite Eq. (5) as S_i^h = exp(q·k/√d + TRB^h(x)); if the additive-after-exponential form is intentional, justify why it behaves as a relative bias and explain the effect on the softmax.
minor comments (5)
  1. [Table 6, rows 15-16] The E.P. column is not filled with explicit True/False labels, so the reader cannot tell which row corresponds to with empty proxies and which to without empty proxies despite the text in Sec. 4.2 describing both conditions.
  2. [Table 8] The table title says Indoor instance segmentation, but the content and metrics (mIoU on ScanNet/ScanNet200 semantic segmentation) describe semantic segmentation; please retitle it.
  3. [Sec. 4.3 and Fig. 9] The caption of Fig. 9 refers to a point-to-point SPA attention map, but the paper only defines point-to-proxy and proxy-to-point SPA; please clarify how the point-to-point map is obtained from the proxy attention components.
  4. [Algorithm 1 and Fig. 6] Algorithm 1 returns (l+r)/2 after max_iter even if the cell count is outside the target range; please state what happens in that fallback case. Also, Fig. 6(b) contains the label TRE Self-Attention with TRE, which appears to be a typo for TRB.
  5. [Appendix A.1] The sentence we evaluated some SOTA models ... as shown in Fig. 8 should refer to Table 8; similarly, a few places use pesade-code instead of pseudo-code.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity: benchmark claims are external comparisons, and design choices are ablations rather than fitted predictions.

full rationale

The central claims in this paper are state-of-the-art results on ScanNet, ScanNet200, S3DIS, SemanticKITTI, nuScenes, and Waymo, obtained by comparing SP2T against independently published baselines such as PTv3, SPoTr, and PointASNL. These are empirical benchmark outcomes, not quantities derived from fitted constants or from the paper's own definitions. The proxy count, association count, TRB table size, and training schedule are selected through ablation studies (Tables 5, 6, 9, and the appendix) and are hyperparameter choices; they do not define the reported accuracy by construction. The paper's novelty claims are supported by ablations that add or remove components on a fixed baseline, which is a standard empirical attribution procedure rather than a circular derivation. The self-citations present in the paper (e.g., references 62-65 by overlapping authors) appear in the introduction, related work, and reference list but are not used to justify the central architecture or to import a uniqueness theorem or ansatz; they are not load-bearing. There is an apparent index inconsistency in the SPA equations (Eqs. 2-3), concerning whether normalization is over proxies or points and whether the output is indexed by point or proxy. This is a potential correctness or notation issue in the description of the attention mechanism, but it does not make the claimed benchmark results circular: the results depend on the implemented code, not on a self-defined equivalence. Accordingly, no circular step can be exhibited, and the paper is self-contained with respect to its external empirical validation.

Assumptions & free parameters 5 free parameters · 3 assumptions · 2 invented entities

The architecture rests on standard attention math and on the geometric assumption that axis-aligned grid vertices capture spatial neighborhoods in sparse point clouds. The tunable hyperparameters are listed. No new physical entities are postulated; the proxy and TRB table are architectural components evaluated only on benchmarks.

free parameters (5)
  • Proxy count M = 160 indoor, 400 outdoor
    Chosen by proxy-number ablation on ScanNet validation (Table 6) and set differently for outdoor datasets in Table 14.
  • Association count k = 8
    Chosen as the number of grid-cell corners and validated by the association-count ablation in Table 6 rows 7-10.
  • TRB table size and strength = size 16, strength 1.0
    Hyperparameters of the table-based relative bias module, listed in Table 14, tuned on the target datasets.
  • TRB input scales = point-proxy 2.5 indoor / 0.2 outdoor, proxy-wise 0.4 indoor / 0.04 outdoor
    Dataset-specific scaling of relative positions before trilinear interpolation; chosen by hand or validation search, not derived.
  • Proxy embedding temperature = 10 indoor, 1 outdoor
    A temperature parameter for sinusoidal position embedding, listed in Table 14 and tuned per dataset.
assumptions (3)
  • domain assumption Increasing group attention window size causes attention dilution and reduces accuracy (PTv3 77.3 mIoU with 1024 points vs 77.1 with 4096 on ScanNet).
    This motivates the entire proxy design and is cited as a known phenomenon in Section 1, but it is not proven for the specific networks used here.
  • domain assumption The eight grid-cell corners give a sufficient sparse proxy neighborhood for each point, i.e., L-infinity KNN with k=8 is adequate for global feature exchange.
    Stated in Section 3.2 and used to define all sparse attention connections; only indirectly tested via the k=4/8/12/16 ablation in Table 6.
  • standard math Sparse softmax computed by map-reduce over association groups is a valid attention normalization that preserves the intended attention distribution.
    Equations 2 and 3 rely on a per-proxy normalization; this is standard softmax grouping but requires care when the similarity includes a learned bias that can be negative.
invented entities (2)
  • Grid-vertex proxies with spatial-wise sampling
    purpose: Compressed global representation stream for point cloud transformers
    A new architectural component whose value is measured only through benchmark accuracy; no external falsifiable handle outside the paper's experiments.
  • Table-based relative bias (TRB) lookup table
    purpose: Provides spatial awareness in sparse proxy attention
    A learned parameterized module; its contribution is shown through ablations but it has no independent predictive power outside the trained models.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SP$^2$T: Sparse Proxy Attention for Dual-stream Point Transformer." pith.science (2026). https://pith.science/paper/FOHCPHLY

@misc{pith2026241211540,
  author       = {Pith},
  title        = {Pith review of: SP$^2$T: Sparse Proxy Attention for Dual-stream Point Transformer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FOHCPHLY}},
  note         = {Machine review of arXiv:2412.11540}
}
abstract

Point transformers have demonstrated remarkable progress in 3D understanding through expanded receptive fields (RF), but further expanding the RF leads to dilution in group attention and decreases detailed feature extraction capability. Proxy, which serves as abstract representations for simplifying feature maps, enables global RF. However, existing proxy-based approaches face critical limitations: Global proxies incur quadratic complexity for large-scale point clouds and suffer positional ambiguity, while local proxy alternatives struggle with 1) Unreliable sampling from the geometrically diverse point cloud, 2) Inefficient proxy interaction computation, and 3) Imbalanced local-global information fusion; To address these challenges, we propose Sparse Proxy Point Transformer (SP$^{2}$T) -- a local proxy-based dual-stream point transformer with three key innovations: First, for reliable sampling, spatial-wise proxy sampling with vertex-based associations enables robust sampling on geometrically diverse point clouds. Second, for efficient proxy interaction, sparse proxy attention with a table-based relative bias effectively achieves the interaction with efficient map-reduce computation. Third, for local-global information fusion, our dual-stream architecture maintains local-global balance through parallel branches. Comprehensive experiments reveal that SP$^{2}$T sets state-of-the-art results with acceptable latency on indoor and outdoor 3D comprehension benchmarks, demonstrating marked improvement (+3.8% mIoU vs. SPoTr@S3DIS, +22.9% mIoU vs. PointASNL@Sem.KITTI) compared to other proxy-based point cloud methods.

Figures

Figures reproduced from arXiv: 2412.11540 by the authors.

Figure 1
Figure 1. The comparison with other Proxy-based Methods. The [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. (a) Overview of SP2T. Followed by PTv3, our model is a U-Net Style Transformer. (b) Architecture of SP2T Layer. Every SP2T layer comprises several SP2T blocks, where both point and proxy must execute four essential operations: local fusion, point-proxy interaction, and global fusion combined with proxy-point interaction. feature of the point F 0 point ∈ R N×C and the proxy F 0 proxy ∈ RM×C and the position of the pr… view at source ↗
Figure 3
Figure 3. Comparison between different grip-based sampling meth [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (6 more)
Figure 6
Figure 6. Figure 6: (a) The architecture of Sparse proxy Attention of point [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 8
Figure 8. Figure 8: Visualization of Point-Proxy and Proxy-Point Attention [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 10
Figure 10. Figure 10: Result Comparison in ScanNet. (a) Ground Truth (b) PTv3 (c) Ours [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: Failed cases in ScanNet. date our model’s test results employing over-segmentation in the final version of the paper. Furthermore, we recom￾mend that future research ensure consistency in the TTA between the validation and test sets or at least make the TTA on the tes…
Figure 12
Figure 12. Figure 12: Visualization of the point-point attention map under FPS-based sampling. The red star represents the current point. [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]
Figure 13
Figure 13. Figure 13: (a) MSE of TRB in different proxy-point interaction [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

69 extracted references · 56 canonical work pages

  1. [1]

    3d semantic parsing of large-scale indoor spaces

    Iro Armeni, Ozan Sener, Amir R Zamir, Helen Jiang, Ioannis Brilakis, Martin Fischer, and Silvio Savarese. 3d semantic parsing of large-scale indoor spaces. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1534–1543, 2016. 2, 6, 14

  2. [2]

    Semantickitti: A dataset for semantic scene understanding of lidar sequences

    Jens Behley, Martin Garbade, Andres Milioto, Jan Quenzel, Sven Behnke, Cyrill Stachniss, and Jurgen Gall. Semantickitti: A dataset for semantic scene understanding of lidar sequences. In Proceedings of the IEEE/CVF international conference on computer vision, pages 9297–9307, 2019. 2, 6, 14

  3. [3]

    The lov ´asz-softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks

    Maxim Berman, Amal Rannen Triki, and Matthew B Blaschko. The lov ´asz-softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 4413–4421,

  4. [4]

    nuscenes: A multi- modal dataset for autonomous driving

    Holger Caesar, Varun Bankiti, Alex H Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Gi- ancarlo Baldan, and Oscar Beijbom. nuscenes: A multi- modal dataset for autonomous driving. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11621–11631, 2020. 2, 6, 13, 14

  5. [5]

    Multi-view 3d object detection network for autonomous driv- ing

    Xiaozhi Chen, Huimin Ma, Ji Wan, Bo Li, and Tian Xia. Multi-view 3d object detection network for autonomous driv- ing. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, pages 1907–1915, 2017. 2

  6. [6]

    Mobile-former: Bridging mobilenet and transformer

    Yinpeng Chen, Xiyang Dai, Dongdong Chen, Mengchen Liu, Xiaoyi Dong, Lu Yuan, and Zicheng Liu. Mobile-former: Bridging mobilenet and transformer. In Proceedings of the IEEE/CVF conference on computer vision and pattern recog- nition, pages 5270–5279, 2022. 2

  7. [7]

    V oxelnext: Fully sparse voxelnet for 3d object detection and tracking

    Yukang Chen, Jianhui Liu, Xiangyu Zhang, Xiaojuan Qi, and Jiaya Jia. V oxelnext: Fully sparse voxelnet for 3d object detection and tracking. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 21674–21683, 2023. 2

  8. [8]

    2-s3net: Attentive feature fusion with adaptive feature selection for sparse semantic segmentation network

    Ran Cheng, Ryan Razani, Ehsan Taghavi, Enxu Li, and Bing- bing Liu. 2-s3net: Attentive feature fusion with adaptive feature selection for sparse semantic segmentation network. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12547–12556, 2021. 6

Show all 69 references
  1. [9]

    4d spatio-temporal convnets: Minkowski convolutional neural networks

    Christopher Choy, JunYoung Gwak, and Silvio Savarese. 4d spatio-temporal convnets: Minkowski convolutional neural networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3075–3084,

  2. [10]

    Pointcept: A codebase for point cloud perception research

    Pointcept Contributors. Pointcept: A codebase for point cloud perception research. https://github.com/ Pointcept/Pointcept, 2023. 13

  3. [11]

    Scannet: Richly-annotated 3d reconstructions of indoor scenes

    Angela Dai, Angel X Chang, Manolis Savva, Maciej Hal- ber, Thomas Funkhouser, and Matthias Nießner. Scannet: Richly-annotated 3d reconstructions of indoor scenes. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 5828–5839, 2017. 1, 2, 6, ...

  4. [12]

    Vision transformers need registers

    Timoth´ee Darcet, Maxime Oquab, Julien Mairal, and Piotr Bo- janowski. Vision transformers need registers. arXiv preprint arXiv:2309.16588, 2023. 1, 2

  5. [13]

    Em- bracing single stride 3d object detector with sparse trans- former

    Lue Fan, Ziqi Pang, Tianyuan Zhang, Yu-Xiong Wang, Hang Zhao, Feng Wang, Naiyan Wang, and Zhaoxiang Zhang. Em- bracing single stride 3d object detector with sparse trans- former. In Proceedings of the IEEE/CVF conference on com- puter vision and pattern recognition, pages 8458...

  6. [14]

    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. 12

  7. [15]

    Agent attention: On the integration of softmax and linear attention

    Dongchen Han, Tianzhu Ye, Yizeng Han, Zhuofan Xia, Siyuan Pan, Pengfei Wan, Shiji Song, and Gao Huang. Agent attention: On the integration of softmax and linear attention. arXiv preprint arXiv:2312.08874, 2023. 1, 2

  8. [16]

    A survey on vision transformer

    Kai Han, Yunhe Wang, Hanting Chen, Xinghao Chen, Jianyuan Guo, Zhenhua Liu, Yehui Tang, An Xiao, Chunjing Xu, Yixing Xu, et al. A survey on vision transformer. IEEE transactions on pattern analysis and machine intelligence, 45 (1):87–110, 2022. 7

  9. [17]

    Mapdistill: Boosting efficient camera-based hd map construction via camera-lidar fusion model distillation

    Xiaoshuai Hao, Ruikai Li, Hui Zhang, Dingzhe Li, Rong Yin, Sangil Jung, Seung-In Park, ByungIn Yoo, Haimei Zhao, and Jing Zhang. Mapdistill: Boosting efficient camera-based hd map construction via camera-lidar fusion model distillation. In European Conference on Computer Visio...

  10. [18]

    Scatterformer: Efficient voxel transformer with scattered lin- ear attention

    Chenhang He, Ruihuang Li, Guowen Zhang, and Lei Zhang. Scatterformer: Efficient voxel transformer with scattered lin- ear attention. arXiv preprint arXiv:2401.00912, 2024. 2

  11. [19]

    Vision transformer with super token sampling

    Huaibo Huang, Xiaoqiang Zhou, Jie Cao, Ran He, and Tie- niu Tan. Vision transformer with super token sampling. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 22690–22699, 2023. 1, 2

  12. [20]

    Perceiver io: A general architecture for structured inputs & outputs

    Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Kop- pula, Daniel Zoran, Andrew Brock, Evan Shelhamer, et al. Perceiver io: A general architecture for structured inputs & outputs. arXiv preprint arXiv:2107.14795, 2021. 1, 2

  13. [21]

    Perceiver: General perception with iterative attention

    Andrew Jaegle, Felix Gimeno, Andy Brock, Oriol Vinyals, Andrew Zisserman, and Joao Carreira. Perceiver: General perception with iterative attention. In International confer- ence on machine learning, pages 4651–4664. PMLR, 2021. 1, 2

  14. [22]

    Pointgroup: Dual-set point group- ing for 3d instance segmentation

    Li Jiang, Hengshuang Zhao, Shaoshuai Shi, Shu Liu, Chi- Wing Fu, and Jiaya Jia. Pointgroup: Dual-set point group- ing for 3d instance segmentation. In Proceedings of the IEEE/CVF conference on computer vision and Pattern recog- nition, pages 4867–4876, 2020. 6, 7, 14

  15. [23]

    Transformers in vision: A survey

    Salman Khan, Muzammal Naseer, Munawar Hayat, Syed Waqas Zamir, Fahad Shahbaz Khan, and Mubarak Shah. Transformers in vision: A survey. ACM computing surveys (CSUR), 54(10s):1–41, 2022. 7

  16. [24]

    Stratified trans- former for 3d point cloud segmentation

    Xin Lai, Jianhui Liu, Li Jiang, Liwei Wang, Hengshuang Zhao, Shu Liu, Xiaojuan Qi, and Jiaya Jia. Stratified trans- former for 3d point cloud segmentation. In Proceedings of 9 the IEEE/CVF conference on computer vision and pattern recognition, pages 8500–8509, 2022. 6, 12

  17. [25]

    Spherical transformer for lidar-based 3d recognition

    Xin Lai, Yukang Chen, Fanbin Lu, Jianhui Liu, and Jiaya Jia. Spherical transformer for lidar-based 3d recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17545–17555, 2023. 6

  18. [26]

    Pointpillars: Fast encoders for object detection from point clouds

    Alex H Lang, Sourabh V ora, Holger Caesar, Lubing Zhou, Jiong Yang, and Oscar Beijbom. Pointpillars: Fast encoders for object detection from point clouds. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12697–12705, 2019. 2, 6

  19. [27]

    Set transformer: A frame- work for attention-based permutation-invariant neural net- works

    Juho Lee, Yoonho Lee, Jungtaek Kim, Adam Kosiorek, Se- ungjin Choi, and Yee Whye Teh. Set transformer: A frame- work for attention-based permutation-invariant neural net- works. In International conference on machine learning , pages 3744–3753. PMLR, 2019. 1, 2

  20. [28]

    Vehicle detection from 3d lidar using fully convolutional network

    Bo Li, Tianlei Zhang, and Tian Xia. Vehicle detection from 3d lidar using fully convolutional network. arXiv preprint arXiv:1608.07916, 2016. 2

  21. [29]

    E-mlp: Effortless online hd map construction with linear priors

    Ruikai Li, Hao Shan, Han Jiang, Jianru Xiao, Yizhuo Chang, Yifan He, Haiyang Yu, and Yilong Ren. E-mlp: Effortless online hd map construction with linear priors. In 2024 IEEE Intelligent Vehicles Symposium (IV), pages 1008–1014. IEEE,

  22. [30]

    Meta architecture for point cloud analysis

    Haojia Lin, Xiawu Zheng, Lijiang Li, Fei Chao, Shanshan Wang, Yan Wang, Yonghong Tian, and Rongrong Ji. Meta architecture for point cloud analysis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17682–17691, 2023. 2

  23. [31]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021. 1, 5

  24. [32]

    Swin transformer v2: Scaling up capacity and resolution

    Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, et al. Swin transformer v2: Scaling up capacity and resolution. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12009–12019, 2022. 1

  25. [33]

    Uedg: uncertainty-edge dual guided camou- flage object detection

    Yixuan Lyu, Hong Zhang, Yan Li, Hanyang Liu, Yifan Yang, and Ding Yuan. Uedg: uncertainty-edge dual guided camou- flage object detection. IEEE Transactions on Multimedia, 26: 4050–4060, 2023. 2

  26. [34]

    Dual- stream network for visual recognition

    Mingyuan Mao, Renrui Zhang, Honghui Zheng, Teli Ma, Yan Peng, Errui Ding, Baochang Zhang, Shumin Han, et al. Dual- stream network for visual recognition. Advances in Neural Information Processing Systems, 34:25346–25358, 2021. 2

  27. [35]

    V oxnet: A 3d convo- lutional neural network for real-time object recognition

    Daniel Maturana and Sebastian Scherer. V oxnet: A 3d convo- lutional neural network for real-time object recognition. In 2015 IEEE/RSJ international conference on intelligent robots and systems (IROS), pages 922–928. IEEE, 2015. 2

  28. [36]

    Mix3d: Out-of-context data augmenta- tion for 3d scenes

    Alexey Nekrasov, Jonas Schult, Or Litany, Bastian Leibe, and Francis Engelmann. Mix3d: Out-of-context data augmenta- tion for 3d scenes. In 2021 international conference on 3d vision (3dv), pages 116–125. IEEE, 2021. 6

  29. [37]

    Fast point transformer

    Chunghyun Park, Yoonwoo Jeong, Minsu Cho, and Jaesik Park. Fast point transformer. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16949–16958, 2022. 1, 2, 6

  30. [38]

    Self-positioning point-based transformer for point cloud understanding

    Jinyoung Park, Sanghyeok Lee, Sihyeon Kim, Yunyang Xiong, and Hyunwoo J Kim. Self-positioning point-based transformer for point cloud understanding. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 21814–21823, 2023. 1, 2, 6

  31. [39]

    How do vision transformers work? arXiv preprint arXiv:2202.06709, 2022

    Namuk Park and Songkuk Kim. How do vision transformers work? arXiv preprint arXiv:2202.06709, 2022. 1

  32. [40]

    Pointnet: Deep learning on point sets for 3d classification and segmentation

    Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 652–660,

  33. [41]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. Advances in neural information processing systems, 30, 2017. 3

  34. [42]

    Point- next: Revisiting pointnet++ with improved training and scal- ing strategies

    Guocheng Qian, Yuchen Li, Houwen Peng, Jinjie Mai, Hasan Hammoud, Mohamed Elhoseiny, and Bernard Ghanem. Point- next: Revisiting pointnet++ with improved training and scal- ing strategies. Advances in neural information processing systems, 35:23192–23204, 2022. 2, 6

  35. [43]

    Language- grounded indoor 3d semantic segmentation in the wild

    David Rozenberszki, Or Litany, and Angela Dai. Language- grounded indoor 3d semantic segmentation in the wild. In European Conference on Computer Vision, pages 125–141. Springer, 2022. 2, 6, 7, 8, 12, 13, 14

  36. [44]

    Mask3d: Mask trans- former for 3d semantic instance segmentation

    Jonas Schult, Francis Engelmann, Alexander Hermans, Or Litany, Siyu Tang, and Bastian Leibe. Mask3d: Mask trans- former for 3d semantic instance segmentation. In 2023 IEEE International Conference on Robotics and Automation (ICRA), pages 8216–8223. IEEE, 2023. 6, 12

  37. [45]

    Semantic scene comple- tion from a single depth image

    Shuran Song, Fisher Yu, Andy Zeng, Angel X Chang, Mano- lis Savva, and Thomas Funkhouser. Semantic scene comple- tion from a single depth image. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1746–1754, 2017. 2

  38. [46]

    Dropout: a simple way to prevent neural networks from overfitting

    Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. Dropout: a simple way to prevent neural networks from overfitting. The journal of machine learning research, 15(1):1929–1958, 2014. 15

  39. [47]

    Multi-view convolutional neural networks for 3d shape recognition

    Hang Su, Subhransu Maji, Evangelos Kalogerakis, and Erik Learned-Miller. Multi-view convolutional neural networks for 3d shape recognition. In Proceedings of the IEEE interna- tional conference on computer vision, pages 945–953, 2015. 2

  40. [48]

    Scalability in perception for autonomous driving: Waymo open dataset

    Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aurelien Chouard, Vijaysai Patnaik, Paul Tsui, James Guo, Yin Zhou, Yuning Chai, Benjamin Caine, et al. Scalability in perception for autonomous driving: Waymo open dataset. InProceedings of the IEEE/CVF conference on computer vis...

  41. [49]

    A brief review of nearest neighbor algorithm for learning and classification

    Kashvi Taunk, Sanjukta De, Srishti Verma, and Aleena Swe- tapadma. A brief review of nearest neighbor algorithm for learning and classification. In 2019 international conference on intelligent computing and control systems (ICCS), pages 1255–1260. IEEE, 2019. 4 10

  42. [50]

    Attention is all you need

    A Vaswani. Attention is all you need. Advances in Neural Information Processing Systems, 2017. 4, 5

  43. [51]

    Dsvt: Dynamic sparse voxel transformer with rotated sets

    Haiyang Wang, Chen Shi, Shaoshuai Shi, Meng Lei, Sen Wang, Di He, Bernt Schiele, and Liwei Wang. Dsvt: Dynamic sparse voxel transformer with rotated sets. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13520–13529, 2023. 2

  44. [52]

    Octformer: Octree-based transformers for 3d point clouds

    Peng-Shuai Wang. Octformer: Octree-based transformers for 3d point clouds. ACM Transactions on Graphics (TOG), 42 (4):1–11, 2023. 6, 12

  45. [53]

    Linformer: Self-attention with linear complexity

    Sinong Wang, Belinda Z Li, Madian Khabsa, Han Fang, and Hao Ma. Linformer: Self-attention with linear complexity. arXiv preprint arXiv:2006.04768, 2020. 1, 2

  46. [54]

    Detr3d: 3d object detection from multi-view images via 3d-to-2d queries

    Yue Wang, Vitor Campagnolo Guizilini, Tianyuan Zhang, Yilun Wang, Hang Zhao, and Justin Solomon. Detr3d: 3d object detection from multi-view images via 3d-to-2d queries. In Conference on Robot Learning, pages 180–191. PMLR,

  47. [55]

    Point transformer v2: Grouped vector attention and partition-based pooling

    Xiaoyang Wu, Yixing Lao, Li Jiang, Xihui Liu, and Heng- shuang Zhao. Point transformer v2: Grouped vector attention and partition-based pooling. Advances in Neural Information Processing Systems, 35:33330–33342, 2022. 1, 2, 6, 7, 13

  48. [56]

    Point transformer v3: Simpler faster stronger

    Xiaoyang Wu, Li Jiang, Peng-Shuai Wang, Zhijian Liu, Xihui Liu, Yu Qiao, Wanli Ouyang, Tong He, and Hengshuang Zhao. Point transformer v3: Simpler faster stronger. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4840–4851, 2024. 1, 2...

  49. [57]

    Pointasnl: Robust point clouds processing using nonlocal neural networks with adaptive sampling

    Xu Yan, Chaoda Zheng, Zhen Li, Sheng Wang, and Shuguang Cui. Pointasnl: Robust point clouds processing using nonlocal neural networks with adaptive sampling. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5589–5598, 2020. 1, 2, 6

  50. [58]

    2dpass: 2d priors assisted semantic segmentation on lidar point clouds

    Xu Yan, Jiantao Gao, Chaoda Zheng, Chao Zheng, Ruimao Zhang, Shuguang Cui, and Zhen Li. 2dpass: 2d priors assisted semantic segmentation on lidar point clouds. In European Conference on Computer Vision, pages 677–695. Springer,

  51. [59]

    Swin3d: A pretrained transformer backbone for 3d indoor scene under- standing

    Yu-Qi Yang, Yu-Xiao Guo, Jian-Yu Xiong, Yang Liu, Hao Pan, Peng-Shuai Wang, Xin Tong, and Baining Guo. Swin3d: A pretrained transformer backbone for 3d indoor scene under- standing. arXiv preprint arXiv:2304.06906, 2023. 6, 12

  52. [60]

    Center- based 3d object detection and tracking

    Tianwei Yin, Xingyi Zhou, and Philipp Krahenbuhl. Center- based 3d object detection and tracking. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11784–11793, 2021. 6

  53. [61]

    Safdnet: A simple and effective net- work for fully sparse 3d object detection

    Gang Zhang, Junnan Chen, Guohuan Gao, Jianmin Li, Si Liu, and Xiaolin Hu. Safdnet: A simple and effective net- work for fully sparse 3d object detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14477–14486, 2024. 4

  54. [62]

    Sparse agent transformer for unified voxel and image feature extraction and fusion

    Hong Zhang, Jiaxu Wan, Ziqi He, Jianbo Song, Yifan Yang, and Ding Yuan. Sparse agent transformer for unified voxel and image feature extraction and fusion. Information Fusion, page 102455, 2024. 2

  55. [63]

    P2ftrack: Multi-object tracking with motion prior and feature posterior

    Hong Zhang, Jiaxu Wan, Jing Zhang, Ding Yuan, XuLiang Li, and Yifan Yang. P2ftrack: Multi-object tracking with motion prior and feature posterior. ACM Transactions on Multimedia Computing, Communications and Applications, 2024. 1

  56. [64]

    Coddiff: Prior leading diffusion model for camouflage object detection

    Hong Zhang, Yixuan Lyu, Tian He, Xuliang Li, Yawei Li, Ding Yuan, and Yifan Yang. Coddiff: Prior leading diffusion model for camouflage object detection. Knowledge-Based Systems, page 113381, 2025. 2

  57. [65]

    Awaretrack: Object awareness for visual tracking via templates interaction

    Hong Zhang, Jianbo Song, Hanyang Liu, Yang Han, Yifan Yang, and Huimin Ma. Awaretrack: Object awareness for visual tracking via templates interaction. Image and Vision Computing, 154:105363, 2025. 2

  58. [66]

    Dspoint: Dual-scale point cloud recognition with high-frequency fusion

    Renrui Zhang, Ziyao Zeng, Ziyu Guo, Xinben Gao, Kexue Fu, and Jianbo Shi. Dspoint: Dual-scale point cloud recognition with high-frequency fusion. arXiv preprint arXiv:2111.10332,

  59. [67]

    Pointweb: Enhancing local neighborhood features for point cloud processing

    Hengshuang Zhao, Li Jiang, Chi-Wing Fu, and Jiaya Jia. Pointweb: Enhancing local neighborhood features for point cloud processing. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 5565–5573, 2019. 2

  60. [68]

    Point transformer

    Hengshuang Zhao, Li Jiang, Jiaya Jia, Philip HS Torr, and Vladlen Koltun. Point transformer. In Proceedings of the IEEE/CVF international conference on computer vision, pages 16259–16268, 2021. 1, 2, 6

  61. [69]

    V oxelnet: End-to-end learning for point cloud based 3d object detection

    Yin Zhou and Oncel Tuzel. V oxelnet: End-to-end learning for point cloud based 3d object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 4490–4499, 2018. 2 11 Indoor Sem. Seg. ScanNet [11] ScanNet200 [43] Methods with same TTA...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.