- Rating
- 3
- Confidence
- 2
- Informativeness
- 3
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper aims to recognize and segment Oracle Bone Inscription (OBI) characters into semantic components. While the goal is to achieve component segmentation, the only available label is whether this component exists in the character. Due to the high annotation cost of pixel-level supervision, this paper proposes a weakly supervised method, HiGAS. HiGAS constructs bottom-up hierarchical undirected graphs at the segment, sub-stroke, and stroke levels for both input characters and components. It first establishes structural correspondences through within-level edge matching, and then propagates higher-level topological priors to enhance low-confidence segment-level responses. Together with structural completeness and hierarchical consistency objectives, the method aims to alleviate topological discontinuity and under-segmentation caused by local feature overfitting under image-level weak supervision.
Critical strengths
- The motivation for weakly-supervised setting is real.
- The results in Table 1 and Table 2 are very impressive. The improvement is large.
- Using loss in a hierarchical way seems to provide much better results.
Critical weaknesses
- There are a lot of manually designed features that lack ablation. For example, for higher-level strokes, there are covered segment count, cumulative and maximum inflection angles, boundary-closure proximity. I am wondering what if we use the naive GNN? Or inversely only use handcrafted features. It is unclear whether the performance mainly comes from the hierarchical reasoning or handcrafted priors.
- The classification loss applies the same image-level presence supervision independently at the segment, sub-stroke, stroke, and fused levels. However, this paper only ablates that this set of loss is necessary without ablating these hierarchical auxiliary losses individually, making it unclear whether supervision at every graph level is necessary, especially since higher-level scores already influence the final prediction through cross-level enhancement.
Questions for authors
In Eq.1, why use attention-like similarity instead of standard cosine similarity?
4 comments
Response to Reviewer 67sC (Part 2/2: Hierarchical Auxiliary Loss Ablation)
Dear Reviewer 67sC,
Below is Part 2/2 of our response, addressing your question regarding the individual contributions of hierarchical auxiliary classification losses (W2).
[W2] Individual Ablation of Hierarchical Auxiliary Losses in
The classification loss applies the same image-level presence supervision independently at the segment, sub-stroke, stroke, and fused levels. However, this paper only ablates that this set of loss is necessary without ablating these hierarchical auxiliary losses individually...
We agree that an overall ablation alone is insufficient to clarify the individual contribution of supervision at each graph level. In Equation (7), the classification loss consists of the fused-level classification loss and three level-specific auxiliary classification losses for :
While keeping the three-level graph structure, cross-level enhancement, and unchanged, we evaluated eight supervision settings: (1) independently removing each of the three auxiliary losses, (2) retaining with only one auxiliary loss at a time, and (3) retaining only without any auxiliary losses:
| Setting | Hit@1 (%) | Hit@3 (%) | Hit@5 (%) | IoU (%) | Prec. (%) | Rec. (%) | F1 (%) |
|---|---|---|---|---|---|---|---|
| HiGAS | 80.97 | 94.07 | 96.77 | 64.35 | 92.65 | 68.92 | 77.58 |
| Without | 80.07 | 93.89 | 96.63 | 59.64 | 87.69 | 64.61 | 72.70 |
| Without | 79.44 | 93.55 | 96.55 | 58.42 | 84.11 | 64.44 | 71.16 |
| Without | 80.83 | 94.03 | 96.53 | 59.60 | 86.76 | 65.44 | 72.87 |
| Fused + segment | 79.36 | 93.08 | 96.24 | 58.62 | 84.94 | 64.84 | 71.83 |
| Fused + sub-stroke | 80.35 | 93.45 | 96.31 | 59.63 | 87.67 | 64.53 | 72.69 |
| Fused + stroke | 79.62 | 93.49 | 96.47 | 58.93 | 86.13 | 64.71 | 71.78 |
| Fused only | 78.61 | 93.50 | 96.36 | 55.94 | 84.56 | 60.95 | 68.96 |
Removing any single auxiliary loss reduces segmentation IoU by 4.71 to 5.93 percentage points, while retaining with only one auxiliary loss improves over using alone (55.94% IoU) but remains below the full model (64.35% IoU). This demonstrates that although cross-level enhancement propagates higher-level scores downward, such interaction cannot fully replace direct classification supervision at individual graph levels. Applying auxiliary losses across all three levels provides complementary supervision that enables the model to better exploit multi-level structural information.
(Updated in App. F, Table 5)
Thanks for the detailed rebuttal. My concerns have been addressed.
Thank You for Your Positive Feedback
Dear Reviewer 67sC,
Thank you for confirming that our rebuttal has addressed your concerns. As the discussion deadline approaches, if you feel that our clarifications and additional ablations strengthen the paper, we would be grateful if you would consider updating your rating or increasing your confidence level in the final evaluation.
Response to Reviewer 67sC (Part 1/2: Feature Ablations & Similarity Design)
Dear Reviewer 67sC,
We sincerely thank you for your positive evaluation of our motivation, hierarchical loss design, and experimental results. We divide our response into two parts: Part 1/2 addresses the ablation on handcrafted attributes and GINE message passing (W1) and the similarity design in Equation (1) (Q1); Part 2/2 addresses the individual ablation of hierarchical auxiliary losses (W2). All corresponding updates are highlighted in blue in the revised PDF.
[W1] Ablation on Handcrafted Attributes and GINE Message Passing
To separately evaluate the contributions of the handcrafted attributes and GINE-based neighborhood message passing, we conducted two additional ablation experiments while keeping the three-level graph structure and cross-level enhancement unchanged:
(Updated in App. F, Table 4)
[Q1] Scaled Dot-Product Similarity vs. Cosine Similarity in Equation (1)
We adopt a scaled dot product after learnable projections in Equation (1) primarily for two considerations regarding node matching and subsequent edge correspondence aggregation:
(Updated in Sec. 3.3.1, Page 5)