- Rating
- 3(+1)
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes AnomalyLoop, a two-stage framework for open-set training-free anomaly generation. The first stage, Anomaly Specific Region Exploration (ASRE), identifies fine-grained candidate anomaly regions. The second stage employs Trajectory Optimal Control (TOC) to optimize diffusion sampling using anomaly detection feedback, with additional semantic consistency and diversity constraints. Experiments on MVTec AD and VisA demonstrate improvements in generation quality, diversity, and downstream anomaly detection performance.
Critical strengths
-
The proposed generation–detection closed-loop framework provides an interesting approach to improving anomaly generation without fine-tuning the pretrained diffusion model.
-
The method achieves promising improvements in generation quality and downstream anomaly detection on MVTec AD and VisA, demonstrating its potential practical value.
Critical weaknesses
-
The proposed ASRE strategy lacks sufficient validation regarding the spatial diversity of selected anomaly regions. Its advantages over simpler random-mask generation strategies remain unclear.
-
The downstream evaluation relies on only 100 generated anomalous images per category from a single normal reference image, raising concerns about instance-specific bias and generalization. Additionally, the lack of generation-time measurements and broader comparisons limits the assessment of practical efficiency.
Questions for authors
-
Could you please clarify whether ASRE produces sufficiently diverse anomaly locations across repeated generations? A comparison of downstream detection performance using ASRE versus random or noise-generated binary masks would help demonstrate its effectiveness.
-
Could you please clarify whether the downstream detectors are trained separately for each category or jointly across categories? If trained separately, would 100 synthetic images per category be sufficient for effective training? If trained jointly, how does this affect the detection performance of individual categories? Furthermore, generating all anomalies from a single normal reference image may introduce instance-specific bias due to limited object appearance diversity. Could you please clarify how this potential issue is addressed?
-
Could you please provide the missing baseline results in Figure 5 and other qualitative comparisons, particularly those marked as N/A? These comparisons appear feasible to complete.
-
Could you please include experimental comparisons with other trajectory optimization or anomaly-aware sampling methods, such as FAST (NeurIPS 2025) or other methods, particularly in terms of downstream detection performance and generation efficiency?
4 comments
Table 7. Inference time per generated image, total training time, and GPU memory usage.
| Method | Total training time (h) | Training GPU memory (GB) | Inference time (s/image) | Inference GPU memory (GB) |
|---|---|---|---|---|
| AnomalyDiffusion (2024) | 158 | 19 | 14 | 13 |
| FAST (2025) | 918 | 29 | 4 | 8 |
| Anomagic (2026) | 6948 | 26 | 3 | 5 |
| AnoStyler (2026) | 0 | – | 15 | 5 |
| AnomalyAny (2025) | 0 | – | 72 | 25 |
| Ours | 0 | – | 39 | 16 |
Table 8. One-shot detection on MVTec AD with an AnomalyGPT-based detector.
| Method | I-AUC | I-F1 | P-AUC | P-F1 | PRO |
|---|---|---|---|---|---|
| AnoStyler (2026) | 90.7 ± 0.5 | 93.4 ± 0.1 | 96.6 ± 0.1 | 62.6 ± 0.2 | 90.1 ± 0.1 |
| Anomagic (2026) | 90.1 ± 0.3 | 93.0 ± 0.2 | 96.7 ± 0.0 | 62.6 ± 0.0 | 91.3 ± 0.2 |
| Ours | 98.2 ± 0.1 | 96.9 ± 0.2 | 97.8 ± 0.1 | 65.9 ± 0.3 | 93.7 ± 0.2 |
Table 15. Per-category segmentation results on MVTec AD under the FAST protocol.
| Category | AnomalyDiffusion mIoU | AnomalyDiffusion Acc | FAST mIoU | FAST Acc | Ours mIoU | Ours Acc |
|---|---|---|---|---|---|---|
| bottle | 76.39 | 83.54 | 86.86 | 90.90 | 87.15 | 95.69 |
| cable | 62.51 | 74.48 | 73.71 | 77.94 | 86.88 | 97.35 |
| capsule | 37.73 | 44.72 | 63.22 | 71.12 | 87.67 | 98.83 |
| carpet | 64.67 | 73.59 | 73.84 | 83.53 | 77.96 | 97.01 |
| grid | 38.70 | 51.82 | 52.45 | 70.70 | 86.52 | 95.48 |
| hazel_nut | 59.33 | 67.48 | 90.81 | 94.79 | 91.38 | 98.10 |
| leather | 56.45 | 62.51 | 66.60 | 74.18 | 84.44 | 98.54 |
| metal_nut | 88.00 | 91.10 | 94.65 | 96.88 | 84.78 | 96.20 |
| pill | 83.21 | 89.00 | 90.17 | 94.07 | 89.30 | 97.47 |
| screw | 38.47 | 49.49 | 49.94 | 57.48 | 88.32 | 98.77 |
| tile | 84.29 | 89.72 | 90.13 | 93.77 | 80.74 | 93.17 |
| toothbrush | 48.68 | 64.41 | 74.98 | 88.63 | 90.39 | 95.49 |
| transistor | 79.27 | 91.74 | 91.80 | 94.50 | 88.05 | 97.47 |
| wood | 60.16 | 74.62 | 78.77 | 86.31 | 85.41 | 96.75 |
| zipper | 65.36 | 72.66 | 72.80 | 84.73 | 88.22 | 98.14 |
| Average | 62.88 | 72.06 | 76.72 | 83.97 | 86.48 | 96.96 |
Q4. FAST comparison and generation efficiency.
Under FAST’s downstream segmentation protocol, AnomalyLoop achieved substantially higher mean IoU and accuracy than FAST on MVTec AD (Table 15). These results demonstrate AnomalyLoop’s advantage over FAST in downstream anomaly segmentation. See Tables 7 and 8 for comparisons with other recent methods in terms of generation efficiency and downstream detection performance.
Thank you for the additional experiments and clarifications, which have addressed most of my concerns, and I will raise my rating to 3.
However, I would encourage the authors to include downstream results using anomalies generated from multiple distinct normal images per category, rather than relying solely on the one-shot setting in the final version. Since normal images are generally abundant in industrial applications, such an evaluation would provide stronger evidence of the method's generalizability.
We sincerely thank you for your positive feedback on our additional experiments and clarifications, as well as for suggesting the use of multiple distinct normal images per category as references for anomaly generation. Following your suggestion, we have added a full-shot experiment on MVTec AD to Section 4.7 of the revised manuscript. For each category, we use all normal training images as references to generate anomalous samples and report the downstream detection results for that category. Given the space constraints in the main text, we report the 2-shot downstream detection results in Appendix G, using two distinct normal reference images per category.
We thank the reviewer for finding our approach interesting and for the positive comments on its promising performance.
Table 3. Controlled ablation results on MVTec AD.
Q1. Downstream detection with ASRE versus random binary masks.
In the w/o ASRE variant in Table 3, we replace the candidate region selected by ASRE with a random binary mask sampled within the object mask. We keep the area budget and subsequent generation settings unchanged and retrain the downstream U-Net for each variant under the same protocol. This random-mask comparison demonstrates the effectiveness of ASRE.
Table 13. Per-category vs. joint training for one-shot detection on MVTec AD.
Table 14. Effect of normal reference selection on one-shot detection on MVTec AD.
Q2. Per-category training, sample size, and instance-specific bias.
Table 2 summarizes the test results across categories. For a fair comparison with prior work, we follow the same protocol and train a separate detector for each category, then test it on that category. We use the AnomalyGPT-based detection framework adopted by AnomalyAny, which extracts visual features with a pretrained CLIP encoder and aligns patch features with text embeddings through a lightweight linear adapter. During training, we keep CLIP frozen and update only the adapter parameters, so the detector does not need to learn visual representations from scratch. Under our 1-shot protocol, 100 synthetic anomaly images per category were sufficient to train the downstream detector effectively. Table 13 reports the test results for a single detector trained jointly across all categories. Its overall performance was lower than that of the per-category detectors, indicating that joint training affects performance on individual categories. We plan to explore how to improve per-category performance under joint training in future work. To assess potential instance-specific bias from using a single normal reference image, we repeated the experiment three times. In each run, we randomly selected one normal reference image per category, regenerated the anomaly samples, and evaluated downstream detection performance. The detection results were similar across the three runs (Table 14), indicating that our framework is insensitive to instance-specific bias in this setting.
Q3. Missing baseline results in Figure 5.
We reproduced AnomalyDiffusion’s generation results on VisA and added the missing qualitative comparison to Figure 5 in place of the entries marked as N/A.