Identification of Botanical Origin from Pollen Grains in Honey Using Computer Vision-Based Techniques

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript used computer vision-based techniques to identify the botanical origin of honey by analyzing pollen grains. The results showed that the proposed techniques can achieve an accuracy of 70% at the rank-1 accuracy and up to 95% at the rank-5, demonstrating the ability of recent advances in computer vision to trace and identify the botanical origin of honey using only microscopy images of the pollen grains. The study sounds interesting as highlighting the potential for the application of AI-based pollen identification systems in the automation of plant traceability processes in agriculture. Overall, some minor points be considered by the authors before it can be recommended for publication.

1. In Abstract, line 9, “To correctly to identify…”, “to” should be dismissed.

2. In Figure 2, should the arrow point from Fig.2C to Fig2D? Please check it again for more clarification.

3. In Figure 5, MobileNet prediction results, why the picture denoted as spin is in shadow, with confidence score of only 0.000?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

In this manuscript, Thi Nhung Le and colleagues identified the botanical origin of honey samples collected in an urban area of Hanoi, Vietnam. I have following comments:

1, For the Abstract, significance of this study should be explained in the keywords.

2, For the key words, Qinghai-Tibetan Plateau should be included.

3, For the introduction, there are duplicated images in the figure 1, please replace.

4, The section 2 related works should be incorporated into Abstract

5, For the Results, error bar should be presented in the Figure 9.

6, For the Discussion section, I would like to see this section could be divided into several subsections and each subsection is properly entitled.

7, For the Materials and methods, randomization in the samplings should be clearly described. Location of experiment sites should be indicated in a map.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

The article’s contribution is not clearly defined. It mentions the implementation of a database, followed by the development of a platform for experts, and suggests the potential for use in process automation. However, much of the pollen extraction process—particularly for larger grains—remains manual, which undermines the proposed automation claims. Figure 1 presents pollen grains, but it is unclear whether the columns represent repeated images of the same grain or independent samples taken before and after a specific process. The figure lacks any accompanying commentary or explanation of the procedure involved. Figure 2 is difficult to interpret and visually cluttered. The presentation could be improved through alternative visualization methods, as the current use of icons, while somewhat descriptive, fails to convey meaningful or relevant information. Figure 3 and Table 1 are related, but their format makes them difficult to navigate. For example, locating Sample 12 requires counting from left to right across the figure, flipping to the table, and then repeating the process when cross-referencing—resulting in a cumbersome and inefficient reading experience. While the use of neural networks is mentioned, it follows standard practice and does not constitute a novel contribution. The most potentially innovative aspect lies in the sample acquisition process for pollen grains, yet this is not given due emphasis in the article. The results presented in the tables are not exceptional and indicate a low level of automation. The computed data is also limited in significance, as the processing is entirely offline. There is no indication of real-time or online analysis that would justify the use of high-performance computational resources. Section 4.2 lacks adequate justification for the metrics used, and there are insufficient references to support them. Given that these metrics form the basis of the discussion in later sections, stronger foundational support is essential. In Section 3.1, the selection criteria for the references are not explained. General characteristics are listed, but specific attributes relevant to the problem at hand are not discussed.

Comments on the Quality of English Language

The manuscript would benefit from language polishing; a peer-reviewed English editing service is recommended to enhance clarity and academic tone.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Authors have addressed my concerns in the revision.

Author Response

We are deeply grateful for your insightful feedback, which provided a solid foundation for our revisions and has significantly enhanced the clarity, depth, and accessibility of our manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript outlines three main objectives: (1) the development of a database, (2) the evaluation of recognition performance using a neural network, and (3) the implementation of an online platform for expert assessment. These objectives are clearly presented in the body of the paper; however, they are not explicitly mentioned in the abstract. I recommend revising the abstract to reflect these goals, ensuring alignment between the initial summary and the full content of the article. This adjustment would provide readers with a more accurate expectation from the outset.

Structurally, the manuscript has improved significantly. The recent corrections have added valuable context to the subject matter, and the overall organization is now clear and coherent. The three stated objectives are effectively developed in the final sections, and the evaluation methods are well-described and properly supported with relevant references.

Figure 2 could benefit from further refinement. Specifically, the removal of icons would enhance clarity, as some graphical elements appear unnecessary and may distract from the intended message.

Finally, the inclusion of additional graphs and references has notably strengthened the manuscript. These enhancements contribute to the rigor and comprehensiveness of the work. Based on the current version, I find no major concerns that would preclude its publication.

Comments on the Quality of English Language

While the English used in the manuscript is generally understandable, a review by a native or expert English speaker is recommended. Such a revision would greatly enhance the clarity and overall quality of the text.

Author Response

Please see the attachment

Author Response File: Author Response.docx