A Hyperspectral Image Classification Method Based on Adaptive Spectral Spatial Kernel Combined with Improved Vision Transformer

Round 1

Reviewer 1 Report

In this manuscript entitled "A Hyperspectral Image Classification Method Based on Adaptive Spectral Spatial Kernel Combined with Improved Vision Transformer" the authors propose a new spectral-spatial kernel combined with the improved Vision Transformer (ViT) to jointly extract spatial/spectral features to complete classification task.

Overall the manuscript is well organized, with a reasonably large reference list in terms of the state of the art. The proposed approach is fairly well presented as well as the tests carried out.

However, in my opinion, the work presented in this manuscript does not seem very original in terms of methodological development. The originality of this work lies essentially in the choice of combination between different modules or processes in order to classify a hyperspectral image with high spectral resolution.

Moreover, according to the results reported in the tables, the classification results obtained are better than the tested literature methods but in my opinion, they are not very significant.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This manuscript proposed a HS classification based on adaptive spectral Spatial kernel combined with improved vision  transformer. The method has the merits and the experiment demonstrates the effectiveness.

1. It is claimed as "re-attention", what is the difference between re-attention and self-attention?

2. In Fig 2, it is claimed that the selective convolution could adjust the receptive field size, but it is not clear how could it adjust.

3. In the experiment, the ablation study is needed to analyze the effectiveness of proposed contribution.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript describes extensively the methodology followed, but there is no discussion of the results in comparison with those obtained by other authors in this field.

On the other hand, the manuscript does not follow the structure set by the journal for the publication of research articles.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

In this manuscript entitled "A Hyperspectral Image Classification Method Based on Adaptive Spectral Spatial Kernel Combined with Improved Vision Transformer", in its revised version, the authors propose a new spectral-spatial kernel combined with an improved vision transformer (ViT) to jointly extract the spatial/spectral features in order to complete the classification task.

 

Overall, this new version of the manuscript is quite well organized, with a reasonably large list of references in terms of state of the art. The proposed approach is fairly well presented as well as the tests performed.

 

However, in my opinion, some parts of this new version still need to be revised:

Review the form of some sections, in particular the section "Experimental Results".

The "Discussions" section includes table 9, which normally should be in the "Experimental Results" section.

Also, I find, in my opinion, that the "Discussions" section can and should be more understood in terms of description of the results and criticism of the tested methods.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The question are addressed in the revision. There is only one more question:

How many training samples for each dataset, how about small sample case?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Authors have improved the manuscript according to the suggestions made.

Author Response

Thank you very much for your Comments and Suggestions.