OCSCNet-Tracker: Hyperspectral Video Tracker Based on Octave Convolution and Spatial–Spectral Capsule Network

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

 

This paper presents a novel hyperspectral video object tracker based on octave convolution and a spatial-spectral capsule network (OCSCNet-Tracker). A spatial-spectral octave convolution module is designed to integrate spatial and spectral information, effectively enhancing information fusion. The proposed spatial-spectral capsule network can establish relationships between different components and targets at various scales, improving target separability. A confidence threshold judgment module is introduced to relocate lost targets using information from the initial frame and adjacent frames, which is a significant contribution. This paper is well-written and well-prepared for Remote Sensing. However, the paper requires minor revisions before a possible publication.

1.     The spatial spectral octave convolution module approach in Section 3.1 of this paper is an existing technology, and I suggest that the spatial octave convolution module should be described simply.

2.Some minor errors should be revised. The formula in lines 248-249 is unlabelled. Line 297 capsule should be plural here.  “vi” should be replaced by “vj” in Equation 13. Line 418 should be the coordinates of the lower right corner of the bounding box. 

3.In Figures 3 and 4, the structures of the spatial octave convolution and the spectral octave convolution are too similar. I suggest that consideration be given to retaining one figure.

4.Figure 5 is not described in the paper, so you can roughly describe the contents of the figure 5. In Figure 5 “sigmod” is changed to “sigmoid”.

5.There are some errors in the experimental section. Are the challenges in Table 1 for BC and not SC from time to time? Line 467 should be replaced with nine baseline methodologies instead of seven. Line 403 data should be 16.2%.

 

Author Response

Attachment please check our response letter.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The author developed OCSCNet-Tracker: Hyperspectral Video Tracker based on Octave Convolution and Spatial-Spectral Capsule Network, and provided a detailed introduction to the model construction process. They conducted experimental comparisons with various methods, comparing Success Rate, Precision, and FPS, offering a comprehensive evaluation. Overall, the performance is quite good and interesting. It is recommended to publish the paper after making some minor adjustments to the details.

(1) It is suggested to elaborate on Equation (5) in detail, along with the explanation of some variables like m and c.

(2) In lines 310 and 312, Fig 8 is referenced incorrectly. It should be Fig. 8.

(3)The Equation in line 28 is missing the equation number.

Author Response

Attachment please check our response letter.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper proposes a method for tracking targets in hyperspectral videos based on octave convolution and spatial-spectral capsule network. The method aims to improve the tracking of selected objects under complex conditions, illumination and scale changes, by integrating spatial and spectral information. To establish the relationships and dependencies between different object components and efficiently extract features from hyperspectral images, octave convolution and capsule networks are used to model the spectral information.

•             The paper is within the scope of the subject matter of the journal and the presented topic could potentially be interesting for the society.

•             Overall, it is a paper with merit. The theory and methods have been validated. 

 

 

The following comments can also be used to improve the review paper quality.

·        Scientific contribution is not explained in detail. The innovation is not highlighted.

·        The references can reflect broadly what has been published in this field.

·        Conclusions should be better stated by a better interpretation of the data.

·        The complexities and problems arising when using the method in scenes with complex backgrounds and rotations are not discussed.

·        It would be useful to describe in more detail possible future directions for improvements and extensions of the method.

·        Conclusions should be better formulated through better interpretation of the data.

·        It would be nice to make a comment comparing with the latest transform models (such as MixFormer), which also show high performance in tracking tasks. In that case, the true performance and promisingness of the proposed method could be evaluated.

 

 

Comments on the Quality of English Language

The language, grammar and stylistics of the text can be improved by:

·        Using shorter and clearer sentences.

·        Avoiding repetition and unnecessary complexity.

·        Improving the structure and logic of the text.

·        Using formal and professional language. 

Inclusion of lists and tables for better visual clarity.

Author Response

Attachment please check our response letter.

Author Response File: Author Response.pdf