You are currently viewing a new version of our website. To view the old version click .
MDPIMDPI
Search all articles
by
  • Zifei Luo1,2,
  • Wenzhu Yang1,2,* and
  • Ruru Gou1,2
  • et al.

Reviewer 1: Anonymous Reviewer 2: Anonymous Reviewer 3: Anonymous

Round 1

Reviewer 1 Report

Following Review should be incorporated in the manuscript

·         Refine overall abstract

·         Mention the limitation in the existing research

·         Add following references in your manuscript

https://ieeexplore.ieee.org/abstract/document/9729883

https://link.springer.com/chapter/10.1007/978-3-031-05752-6_1

https://link.springer.com/chapter/10.1007/978-3-030-97113-7_5

·         Add more information about the proposed methodology

·         What is the main limitation of your proposed study

 

·         Refine your conclusion

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Thanks for your paper.

These are my corrections and suggestions?

Row 85: To retain -> to maintain? I think it’s better to change this phrase.

Related works: In my opinion, it is useful to talk about The Canny Edge Detector and the Hough Transform

Row 160: How we can define the complexity of an image?

Row 162 Insert a space between © and different scales

Equation 1: It needs to explicit: I_G, I_R, I_B. How are they defined?

Row 179: Remove points after (1) and (2).

Row 192: Insert a space after (b)

Row 259 : Remove “:”

Row 279: Remove “:”

Row 307: Insert a space after (a) and (b)

 

And other minor typos to correct.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper "TransAttention U-Net for Semantic Segmentation of Poppy" is devoted to the recognition of the illegal cultivation of opium poppy based on drone aerial photography. The author's impact consists of using special approaches to image processing, an improved TAU-NET network, and quality UAV-images. The obtained metric's results confirm the effectiveness of the author's method. Dice-score is 0.77, and it is the best result in comparison with U-NET and DeepLabV3+.

A few observations:

1. Justifying the choice of a basic U-Net architecture is not enough. More information should be given about the limitations of other modern models suitable for the task, such as the YOLO and SSD.

2. Did you provide any experiments to determine optimal training parameters? This should be clarified in the article.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf