All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Dear Authors,
Your paper has been accepted for publication in PEERJ Computer Science. The comments of the reviewers who evaluated your manuscript are included in this letter. I ask that you make minor changes to your manuscript based on those comments, before uploading final files. Thank you for your fine contribution.
[# PeerJ Staff Note - this decision was reviewed and approved by Xiangjie Kong, a PeerJ Section Editor covering this Section #]
no comment'
no comment'
no comment'
Dear Authors
Based on the reviewers' reports, your paper requires significant revisions before it can be considered for publication in PEERJ Computer Science.
The major issues are listed in what follows:
1) It is necessary to explain how the main innovative component of this paper, the AIoU loss, is effectively integrated into the object detection processes described in Sections 1.2, 1.3, and 1.4 and why it performs better than other methods.
2) The experimental results could be clearer and more sufficient to show the proposed method's main advantages and contributions. Some additional experiments need to be conducted to strengthen the conclusion.
Furthermore:
1) The sentences and grammar of the manuscript should be carefully checked and revised.
2) Some figures need to be clearer and need to be modified. Tables should be normalized.
This paper analyzed the original BBR loss and proposed AIoU-v1 loss, and designed AIoU-v3 with adaptive dynamic non-monotonic FM.
1. The sentences and grammar of the manuscript should be carefully checked and revised.
2. Some figures are not clear enough and need to be modified. Tables should be normalized.
3. You should be more standardized and reasonable to summarize and review the relevant work. Some related works should be discussed in this paper: --Active learning for deep tracking, --AIoU: Adaptive bounding box regression for accurate oriented object detection, Self-supervised Deep Correlation Tracking
4. The experimental results are very unclear and not enough to show the main advantages and contributions of the proposed method.
5. Experiment validations are not convincing. Some additional experiments need to be conducted to make its conclusion stronger.
no comment
6. Please consider discussing and analyzing the limitations of the proposed method.
The content of the article is mostly understandable. A sufficient background is provided on the loss function, which is the primary topic of the paper. Figures and tables are professionally structured. However, the captions of all figures are very short and do not describe the content of the figures. There are enough experiments and results.
The authors have described the limitations of existing loss functions for bounding box regression and compared them mathematically and numerically. However, following different versions of loss functions was difficult, e.g., v1, v2, etc. Therefore, I recommend putting together a table that mentions a list of loss functions, their formula, and their different versions.
1. I could not find the definition of FM.
2. Which loss function is defined by Eq 3: GIOU or DIOU?
3. Which model is used for regression for the simulation experiments in Sec. 3.1?
4. What do you mean by final error?
5. What does the suffix "-de" signify? Does it indicate the decoupling of height and width from the denominator of the loss function?
6. Does the IOU threshold for filtering the samples (low and high-quality) depend on the dataset used for training the model?
7. Highlight the best model in bold color in each result table.
The authors conclude their findings and compare with state-of-the-art methods.
see attached file
see attached file
see attached file
see attached file
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.