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.
The manuscript has been accepted for publication. Thank you for your contribution.
[# PeerJ Staff Note - this decision was reviewed and approved by Jyotismita Chaki, a PeerJ Section Editor covering this Section #]
Please address all requests and comments of the reviewers in detail.
**PeerJ Staff Note:** Please ensure that all review, editorial, and staff comments are addressed in a response letter and that any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
This paper is based on YOLO11 optimization for lightweight and accurate plant detection in UAV aerial imagery. Thus, this paper is directly related to the theme of this journal.
Overall, the paper is organized properly; the concept and future research directions are extensively explained. So, the paper is accepted after following minor changes:
1. The problem of paper and motivation is not clear in the introduction
2. Contribution of the paper is not given in bullets
3. Comparison of current research is not given with previous work
4. Figure colors are light and not clear, so add dark colors for a clear view
5. The paper contains a few grammar mistakes that should be corrected in the final version.
6. Only a few references are added in the paper; read a few latest papers related to this paper.
Experiment details are given in detail for the reader to easily understand.
-
It's a very good paper overall, presenting a new combination of trimmed-off known architecture elements of the YOLO family. It shows that optimising it by taking out elements that seem redundant for some applications makes it more efficient computationally without loss of performance.
General comments on the paper form that need to be addressed:
-What the authors call “a significant enhancement in performance” is overstated, as from the results, the performance enhancements are mostly small, some marginal, and I some cases actually underperforming slightly. The various places where this is said should be rewritten with more objective language.
- The main value of this work is in an optimised architecture with minimal redundancy. Even though the difference may not be enough to use lower performance computers, such I IoT, as the authors suggest, but when scaling up, small savings accumulate and can make a big difference.
Specific comments on content and analyses that need to be addressed:
- The dataset size is still very limited: while Plant4 is bigger than its components, ~1.3k images are still a very small sample for any deep learning model. This raises questions about generalisation in larger datasets, or even just any datasets outside the authors’ own. The authors disregard the need to comment on this by claiming that the dataset is significantly larger than, but, as stated in this comment, that is not the case. They even failed to consider de unbalanced nature of the dataset, as more than half of the samples are from one class of 4 (MTDC-UAV), and in the validation set, this becomes over 60%.
- While training from scratch ensures fairness, for such small datasets, it is imperative to start with pretrained models or use transfer learning; at least a comparative experiment should be done to validate their claims that starting from scratch works better even in this case.
- Again, given the low number of samples used (and especially starting from scratch as stated in lines 259-260), the paper requires an explicit analysis of borderline cases—e.g., occlusion, noisy images, or background clutter.
It is a very good paper, with nothing to be addressed, as it is covered in the general comments above.
Comments on the raw data that need to be addressed:
- The described dataset is not available; the link they provide is reachable, but the actual data is not there. Nor the complete software (it seems the final version was not committed to GitHub).
- The annotated merge Plant4 data are not found either, and their comment is confusing, so we researched both repositories (the author's repository and the original paper’s one), and the annotations are not in either one.
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.