Optical Remote Sensing Image Cloud Detection with Self-Attention and Spatial Pyramid Pooling Fusion
Round 1
Reviewer 1 Report
Comments for author File: 
Comments.pdf
Author Response
Thank you very much for your suggestion. We have provide a point-by-point response to your comments, please view the upload files.
Author Response File: Author Response.pdf
Reviewer 2 Report
The present paper proposes an improved deep-learning based method for cloud detection in optical images.
Generally speaking, the paper is well organised and interesting with nice results. Although most sections of the paper are generally written in a good English, the abstract and the introduction need an extensive revision of the syntax and of punctuation since several parts are almost not readable at all.
General comments:
11) In the Introduction, there is a need to insert further references to support some statements given therein (e.g. page 2, line 63; page 2, line 72).
22) As far as the novelty of the paper is concerned, further emphasis should be given to the main differences with the state of the art, with reference to the type of data and sensors used in previous works. This implies to extend the state of the art given in the Introduction.
33) In the Results section, section 3.2, it is poorly described the method used to generate the ground truth images (‘manual outlining’).
44) Figure 4 and figure 5: it is previously stated that the other methods include also shadow detection (page 8, line 283 and ff.). In the mentioned figures (and especially in figure 5) it seems there is still residual effects of this on the fact that there are several blue areas corresponding to incorrectly detected pixels (that in the true-color image seems to be associated to shadows)). Please comment on this.
Specific comments:
a. Table at page 8 needs to be formatted differently: the description of the principles the methods are based on is not clear (each description is not clearly separated from the following one).
b. Table 5 needs as well reformatting. It is not easily readable. Spell also journal acronyms.
c. Figure 6: improve the labelling of the sub-figures and relevant caption in order to identify ‘who is who’ more easily.
For all these reasons, in the Referee’s opinion the manuscript needs minor revisions.
Author Response
Thank you very much for your suggestion. We have provide a point-by-point response to your comments, please view the upload files.
Author Response File: Author Response.pdf
Reviewer 3 Report
- The manuscript focused on optical remote sensing image cloud detection with self-attention and spatial pyramid pooling fusion.
- Did you consider using higher resolution optical remote sensing data such as Sentinel-2 with 10m spatial resolution, or even higher resolution (non-commercial data)?
- Which time (from year A to year B) did you consider for the acquisition of the remote sensing data? and did you consider the impact of seasonality, as well as the impact of geographical location and topography, e.g the cloud in coastal areas, are different than in agriculture or forest area?
- could you show the location of the study area to show the surrounding topography and land cover/land use features to understand the source and distribution of clouds in the study area?
Author Response
Thank you very much for your suggestion. We have provide a point-by-point response to your comments, please view the upload files.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
Accept in present form
