Next Article in Journal
Cybersecurity Analytics for the Enterprise Environment: A Systematic Literature Review
Next Article in Special Issue
AI-Powered Gamified Scaffolding: Transforming Learning in Virtual Learning Environment
Previous Article in Journal
A Low-Carbon Economic Scheduling Strategy for Multi-Microgrids with Communication Mechanism-Enabled Multi-Agent Deep Reinforcement Learning
 
 
Article
Peer-Review Record

HyADS: A Hybrid Lightweight Anomaly Detection Framework for Edge-Based Industrial Systems with Limited Data

Electronics 2025, 14(11), 2250; https://doi.org/10.3390/electronics14112250
by Xingrao Ma 1, Yiting Yang 1, Di Shao 2,*, Fong Chi Kit 1 and Chengzu Dong 1,*
Reviewer 1: Anonymous
Reviewer 2:
Reviewer 3:
Reviewer 4:
Electronics 2025, 14(11), 2250; https://doi.org/10.3390/electronics14112250
Submission received: 30 April 2025 / Revised: 22 May 2025 / Accepted: 27 May 2025 / Published: 31 May 2025

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper "HyADS: A Hybrid Lightweight Anomaly Detection Framework for Edge-Based Industrial Systems with Limited Data" proposes a lightweight anomaly detection framework that combines LBP/HOG features, autoencoders (U-Net) and PatchCore methods. The research results have certain value and application potential in the field of industrial anomaly detection. However, the article still needs to be improved in terms of method details, experimental analysis and result presentation. Below, I have outlined specific concerns and suggestions for revision that I believe will enhance the clarity, rigor, and presentation of the research.

 

1.In the method section, clearly explain the specific implementation methods of real-time performance (40-45 FPS), such as model lightweight design or hardware optimization strategy.

2.Supplement the specific dynamic update mechanism of the memory bank, and clarify the methods of sample update and diversity guarantee (see Section 3.3).

3.In the method description (Section 3.1), clearly explain the specific basis for the selection of LBP and HOG feature parameters, such as experimental verification of parameters or reference sources.

4.Detailed description of the specific method of weight determination in the adaptive fusion strategy (Section 3.4), such as whether it is determined by cross-validation or experimental parameter search.

5.In the experimental analysis section (Section 4), add classification analysis and comparison of the detection results of defects of different types or sizes to highlight the advantages and limitations of the method.

6.In the discussion section, clearly point out the specific industrial fields or application scenarios for which this study is suitable, such as precision manufacturing, electronic device detection, or steel surface detection.

7.In the conclusion section, elaborate on the research directions that may be explored in the future, such as the model adaptive update mechanism or in-depth application research in specific industrial fields.

Author Response

Dear reviewer,

Please check the detailed review report in the attached PDF. Thank you!

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This work proposed a Hybrid Anomaly Detection System (HyADS) for edge-based industrial defect detection, which integrates three modules (a feature extractor, a normal pattern reconstructor and an adaptive patch matching module). There are some issues that should be addressed in this work.

  1. The cites of some references are missing.
  2. The related works should include more up-to-date related works, and discuss them in more depth instead of only listing brief introductions for their work.
  3. More reasons why choosing the technologies used by HyADS must be illustrated.
  4. The figures, especially figure 1, should be improved.
  5. The focused problem should be stated in detail.
  6. The method should be illustrated in combination with the problem.
  7. More comprehensive results and evaluations are needed.

Author Response

Dear reviewer,

Please check the detailed review report in the attached PDF. Thank you!

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper proposed a Hybrid Lightweight Anomaly Detection Framework for Edge‑Based Industrial Systems with Limited Data, and the structure of this draft is clear. However, there are some revisions before a possible publication. Below are the reviewer's comments and suggestions aimed at enhancing the clarity, rigor, and overall quality of the manuscript.

  1. Before diving into individual modules, there is no dedicated subsection that clearly outlines the end‑to‑end HyADS pipeline. It is recommended to add a “Framework Overview” section with a concise narrative of the data flow.
  2. The claimed FPS speed lacks direct comparison with competing algorithms. Please provide a runtime comparison table on identical hardware to substantiate the speed‑accuracy trade‑off.
  3. An ablation study isolating the contributions of (a) the texture extractor, (b) the U‑Net autoencoder, (c) the PatchCore module, as well as pairwise fusions and the full HyADS pipeline, is missing.
  4. Sections 3.1 and 3.3 are overly detailed, obscuring the novel adaptive fusion strategy and segmentation head design. It is suggested to condense background information into concise paragraphs with appropriate citations and expand the description of your own contributions.
  5. The comparative experiments omit recent SOTA methods, and the choice of baselines is not sufficiently representative. Please include at least two additional 2024–2025 algorithms in both detection and segmentation tasks to ensure fair and comprehensive comparison.
  6. The Conclusion (Section 6) outlines future work but does not acknowledge the limitations of HyADS itself.

 

Author Response

Dear reviewer,

Please check the detailed review report in the attached PDF. Thank you!

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The article proposes HyADS, a hybrid lightweight anomaly detection framework integrating classical texture descriptors (HOG, LBP), U-Net autoencoder, and PatchCore-based local anomaly detection, targeting edge-based industrial systems with limited data. The article has gaps and needs the following improvements, namely:

1 - Authors should clarify and justify claims about edge deployment; current experiments were conducted on an RTX 4080 GPU, which is not representative of edge devices.
2 - They should consider including results or at least a discussion on realistic edge hardware, such as NVIDIA Jetson Nano, Xavier NX, or similar.
3 - Try to expand the "Related Work" section to include newer and more relevant methods, particularly DRAEM and PaDiM, which are the current state of the art in anomaly detection.
4 - Add an ablation study to clearly quantify the individual contributions of each module (HOG+LBP, U-Net, PatchCore).
5 - Potential limitations related to model scalability and computational overhead in large-scale industrial scenarios are also missing.

Author Response

Dear reviewer,

Please check the detailed review report in the attached PDF. Thank you!

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The revised version solved the problems

Reviewer 2 Report

Comments and Suggestions for Authors

No comment

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have addressed all my comments and I recommend to publish this manuscript in present form.

Back to TopTop