- Rating
- 2
- Confidence
- 2
- Informativeness
- 3
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The paper proposes the DVD–CBG–OW-PC framework for generalized few-shot 3D segmentation. It leverages dual VLMs and support samples to extract reliable novel anchors, preserves high-confidence base regions, and then recovers missing structures via ddCRP and component-level posterior inference. The authors further extend the framework to partial/zero-support unknown-instance recovery and report performance improvements on four indoor 3D benchmarks.
Critical strengths
-
Shift from point-level pseudo-label refinement to structured novel-instance recovery. OW-PC unifies local spatial/feature connectivity, base/background counterfactuals, and support-conditioned component evidence within a coherent component inference formulation. Since ddCRP is inherently suited to represent observation-dependent partitions with a variable number of components, this modeling paradigm aligns well with the goal of recovering missing instances from query structures.
-
There is substantial positive evidence under the standard GFS setting. In Table 1, compared with GFS-VL, the proposed method achieves improved novel mIoU and HM across four datasets and two shot configurations. Table 7 further presents 1-shot results with five random seeds, demonstrating that the main standard-setting results do not strongly rely on a single random run.
Critical weaknesses
- The claim of support-free novel-class discovery lacks solid evidence and has evaluation issues. Under zero-support settings, the model can only detect individual unknown instances but cannot automatically group them into distinct semantic classes. This is unknown-instance discovery rather than true novel-class discovery. Besides, Table 4 uses inconsistent GT denominators across methods, and it is unclear how baseline methods run without supports. The conclusion is not fair. The authors should recompute metrics on the identical GT instance set. Either rephrase the claim as unknown-instance discovery, or show the model can separate different novel classes.
- Ablation results for core modules are inconsistent across tables, making module contributions hard to verify. For ScanNet 1-shot, the full model and DVD-only variant show obvious discrepancies in Tables 1/3/7/10. The paper does not clarify whether these differences come from different VLMs, random seeds, checkpoints or preprocessing. Since the paper claims individual module benefits, experiments must be standardized. Ideally, provide progressive ablation results under the same seed.
- LARM also studies generalized few-shot point cloud segmentation with language priors and structured matching. The authors need to explain the fundamental difference between OW-PC’s query-side structural discovery and LARM’s prototype/relation enhancement.
Questions for authors
1.Why does ScanNet200 in Table 4 have 9,740 GT instances for attMPTI/GW versus 8,029 for GFS-VL/Ours? Can you report [email protected] and mean-best-IoU for all methods on the identical GT subset using existing predictions?
2.What inputs are provided to attMPTI, GW and GFS-VL in zero-support tests? Are novel class names, text prompts, support prototypes and novel semantic logits removed for these baselines identically to our method?
3.For ScanNet 1-shot results in Tables 1, 3, 7 and 10, what seeds, registration settings, VLM pairings and checkpoints were used for each? Specifically, what causes the large gap between the two “DVD” results?
4.Is the predicted novel-category count in Table 2 calculated per scene, over the full validation set, or deduplicated across registration runs? Under , is there any GT-free procedure to group separate components into distinct novel semantic classes?
6 comments
======================== PART 2/3 ========================
W3 : Conceptual comparison with LARM
LARM enhances category prototypes using LLM-generated descriptions and inter-class hyper-relations to improve point classification within a fixed, support-defined vocabulary (Li et al., ICCV 2025, Sections 3.3–3.4). In contrast, OW-PC infers query-side object formation and verification: component evidence shapes the spatial partition (Eq. 18), Eq. (19) verifies completed components against competing explanations, and the unknown hypothesis enables recovery in zero-support (Table 4) and partial-support (Table 5) regimes that LARM does not evaluate.
Because LARM's code is not yet public, we could not execute it directly; Table R2 places its published 1-shot numbers alongside our seed-42 runs (Appendix Table 8) with each protocol recorded.
Table R2: Reported 1-shot results. Base, Novel and HM are percentages.
| Dataset | Method and protocol | Base | Novel | HM |
|---|---|---|---|---|
| S3DIS | LARM (7 base / 6 novel, Area 6 test) | 75.68 | 44.52 | 56.04 |
| S3DIS | Ours (10 base / 3 novel, local split, seed 42) | 60.35 | 33.85 | 43.18 |
| ScanNet | LARM (14 base / 6 novel, 1,201/312 scenes) | 42.71 | 21.50 | 28.58 |
| ScanNet | Ours (13 base / 6 novel, GFS-VL split, seed 42) | 78.44 | 51.92 | 62.48 |
The protocols differ in class splits (LARM uses 7/6 on S3DIS and 14/6 on ScanNet, whereas we follow the 13/6 GFS-VL ScanNet split and our local 10/3 S3DIS split); only the ScanNet novel classes coincide across protocols. We will add a like-for-like comparison under a shared split and sampling protocol once LARM's implementation is released.
Q2 : Inputs to the zero-support baselines
Every method receives only coordinates and frozen point features (Table R3). Novel names, prompts, support prototypes, and semantic logits are excluded from discovery (GFS-VL's internal semantic outputs are discarded), GT labels enter only post-hoc evaluation, and all adapters return anonymous instance components.
Table R3: Inputs passed to discovery by each adapter.
| Method/adaptation | Novel names/prompts | Novel support | Novel logits | Feature source |
|---|---|---|---|---|
| attMPTI adaptation | No | No | No | attMPTI DGCNN |
| GW adaptation | No | No | No | GW DGCNN |
| GFS-VL adaptation | No | No | No | OpenScene SparseUNet |
| Ours | No | No | No | Mosaic (DVD pair) |
======================== PART 3/3 ========================
Q4 : Novel-category count and GT-free grouping
Table 2 reports a per-run, dataset-level count averaged over five runs: , where counts final anonymous category groups pooled across validation scenes in run (Appendix A.5). Because groups are inferred without a supplied class count, can exceed the GT vocabulary size (e.g., S3DIS's reflects over-splitting of the three true categories). Note that Table R4 counts clusters from the shared OpenScene-feature decoder, so it is not directly comparable to Table 2.
Table R4: Support-free grouping with cached OpenScene features and the legacy feature-only decoder; cosine/average linkage at threshold 0.25, -means with . Grouping is transductive.
(a) Novel-instance recovery and semantic grouping
| Category assignment | Recall$\uparrow$ | Coverage$\uparrow$ | ARI$\uparrow$ | NMI$\uparrow$ | ACC$\uparrow$ |
|---|---|---|---|---|---|
| One category per instance | 25.52 | 16.77 | 0.00 | 48.41 | 7.32 |
| One category for all instances | 25.52 | 16.77 | 0.00 | 0.00 | 42.68 |
| K-means (K=99) | 25.52 | 16.77 | 91.77 | 88.98 | 91.46 |
| Ours (OW-PC + semantic grouping) | 25.52 | 16.77 | 92.89 | 91.85 | 92.68 |
(b) Category granularity and base-class mixing
| Category assignment | All clusters | Novel-touching clusters | Mixed base/novel clusters |
|---|---|---|---|
| One category per instance | 125120 | 82 | 0 |
| One category for all instances | 1 | 1 | 1 |
| K-means (K=99) | 99 | 10 | 5 |
| Ours (OW-PC + semantic grouping) | 99 | 11 | 4 |
Note: All eligible components enter the clustering, including base objects; a component needs at least 20 sampled points and a nonzero pooled feature to receive an assignment. Panel (a): Recall is the support-free recall of Table 4 of the manuscript on the same eligible instances, so it is shared across rows; Coverage is the fraction of eligible novel GT with a category-eligible match; ARI, NMI and Hungarian-matched ACC use the matched novel instances. Panel (b): novel-touching clusters contain a prediction matched to novel GT, mixed clusters also one matched to base GT. All scores are percentages.
For GT-free grouping without support, we cluster pooled component features using average linkage (following the GFS-VL feature protocol) and stop merging at a threshold calibrated on held-out base labels and frozen before novel evaluation—without using novel GT, names, prompts, prototypes, or a supplied class count. On all 312 ScanNet validation scenes (Table R4), this grouping outperforms -means on ARI ( vs. ), NMI ( vs. ), and ACC ( vs. ) with fewer mixed base/novel clusters ( vs. ), documenting category granularity while leaving semantic class separation outside its scope.
======================== PART 1/3 ========================
Thank you for carefully reading our paper and for your thoughtful comments.
W1-1 : Claim scope and unknown-instance wording
With labeled support, our method assigns named novel-class labels (Appendix Table 10 reports 57.09% novel-class mean recall vs. 44.98% for GFS-VL on ScanNet200 1-shot); without support, outputs remain anonymous instances, whose grouping is evaluated separately in Q4.
We have revised the title, Abstract, Sections 1 and 5 to state unknown-instance discovery.
W1-2 & Q1 : Inconsistent GT denominators in Table 4
We apologize for the coding oversight. The ScanNet200 denominator mismatch (8,029 vs. 9,740) stemmed from filtering
semantic > 0in the original GFS-VL and Ours runs, which excluded all 1,711wallinstances (ID 0).Table R1: ScanNet200 instance recovery. Recall at IoU 0.50 and mean best IoU are percentages; parenthetical counts are matched/GT.
Retaining
wallaligns all four methods on the identical 9,740-instance domain across the same 312 scenes (Table R1), where GFS-VL recovers 1,977 instances (20.30%, 30.71% mean best IoU) and Ours recovers 2,029 (20.83%, 33.34% mean best IoU). Similarly, on ScanNet++ all rows now use the common 3,186-instance domain (GFS-VL 17.55%, 559/3,186 vs. Ours 20.28%, 646/3,186), and Table 4 now uses a single denominator per dataset (6,240 for ScanNet, 3,186 for ScanNet++, and 2,339 for S3DIS). Every baseline runs without support because its adapter receives only coordinates and frozen point features (audited in Q2).We have corrected the support-free comparison in Section 4.5 (Table 4) and Appendix D.5, and have clarified the decoder inputs, annotation vocabularies, sampling rules, and valid instance IDs in Appendices C.1, C.3, C.4, and C.7 of the revised manuscript.
W2 & Q3 : Seeds, checkpoints and differing module configurations
The discrepancies stemmed from training seeds and module configurations. Every single-run entry (including the progressive ablation in Table 3) now uses seed 42, registration
regis1, and checkpointmodel_last.pth, while Table 1 and Appendix Table 8 report five-seed means (seeds 42, 43, 123, 44, and 45). Consequently, Table 3 now forms a same-seed progression whose full-model ScanNet 1-shot entry is Base 78.44, Novel 51.92, and HM 62.48 (versus the five-seed mean of 78.65, 51.18, and 62.00 in Table 1).Regarding the VLM pairing, Appendix Table 11 compares the two full pipelines and Appendix D.6 reports single-VLM baselines and both OpenScene–Mosaic role assignments. The two "DVD" entries measure different configurations: Table 3's DVD-only row disables CBG and OW-PC on seed 42, whereas Appendix Table 11 retains both modules and compares full pipelines, combining module configuration with teacher choice.
We have standardized all experiments and clarified this convention in Appendix C.2 of the revised manuscript.